University of Ghana http://ugspace.ug.edu.gh FACTOR ANALYSIS OF CUSTOMER PREFERENCE FOR MOBILE PHONE NETWORK (A CASE STUDY OF CAPE COAST POLYTECHNIC) BY CHARLES MENSAH (10513132) THIS THESIS IS SUBMITTED TO THE UNIVERSITY OF GHANA, LEGON IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE AWARD OF MASTER OF PHILOSOPHY DEGREE IN STATISTICS. SEPTEMBER, 2015 University of Ghana http://ugspace.ug.edu.gh DECLARATION Candidate’s Declaration This is to certify that, this thesis is the result of my own research work and no part of it has been presented for another degree in this University or elsewhere. ……………………………. …………………………… CHARLES MENSAH DATE (10513132) Supervisors’ Declaration We hereby certify that this thesis was prepared from the candidate‟s own research work and supervised in accordance with guidelines on supervision of thesis laid down by the University of Ghana, Legon. ……………………………..……. …………………………… PROFESSOR O.A.Y. JACKSON DATE (SUPERVISOR) ……………………………..……. …………………………… PROFESSOR J.B. OFOSU DATE (CO- SUPERVISOR) ……………………………………. ……………………………… MR. C. A. HESSE DATE (HEAD OF DEPT.) i University of Ghana http://ugspace.ug.edu.gh ABSTRACT This research tries to determine the “hidden” factors which ultimately influence the choice of mobile phone network in Cape Coast Polytechnic (study area). Principal Component method of Factor analysis is used to achieve the set objectives. There are six mobile phone network providers (MTN, Vodafone, Tigo, Airtel, Globacom and Expresso) in Ghana. All these operators try to improve their marketing strategies in order to attract more customers or subscribers to increase their market share. A total of 500 respondents were drawn from students, teaching staff and administrative staff in Cape Coast Polytechnic by proportional allocation. There is no restriction in age and gender but it is required that a respondent belongs to one of these three groups. The study made use of research instrument in order to measure attributes of the networks. A 14 item likert scale was used with 5 levels of agreement in the questionnaire. The study shows that more male respondents participated in the study mostly are in the age group of 18-24 years old. Cronbach‟s alpha shows that the data collected is consistent (reliable), while the Kaiser Meyer Olkin test and Bartlett‟s test of sphericity show significant results that factor analysis is appropriate in the data gathered. Therefore, factor analysis is applicable. It is found that “long time usage” is the most important attribute followed by “wider coverage” and “good advert”. Surprisingly, “lower tariff” and “games of chance” happened to be the less important attributes. The most regularly used mobile phone network in Cape Coast Polytechnic is MTN followed by Tigo. Vodafone is the third most regularly used network, while Expresso is the least patronized network. Three factors were extracted, factor one is social responsibility factor (or customer care factor), factor two is reception benefit factor and factor three is relationship benefit factor. These three (3) factors identified, best summarizes the people‟s choice of mobile phone network in Cape Coast Polytechnic. ii University of Ghana http://ugspace.ug.edu.gh DEDICATION I dedicate this thesis to the Almighty God for His grace, protection and guidance. Also, to my parents, Opanyin Kwaw Badu and Madam Ama Ayipeh (all of blessed memory) and my son, Yaw Atta Mensah. iii University of Ghana http://ugspace.ug.edu.gh ACKNOWLEDGEMENT My first thanks go to the Almighty God for His protection, guidance and for giving me wisdom, knowledge and understanding all these years that I have passed through the academic ladder and for the great things He is about to do in my life. My very special thanks go to Professor O.A.Y. Jackson (supervisor) and Professor J.B. Ofosu (co-supervisor) who guided me like a mother guiding her child to take a first step. To my family, I say thank you for your encouragement, support and advice even when all hopes were lost, you kept me going. My greatest thanks go to all the lecturers of the Department of Mathematical Sciences of Methodist University College Ghana and Department of Statistics of University of Ghana, Legon, as well as my MPhil (statistics) colleagues. Finally, my heartfelt gratitude go to the individuals in Cape Coast Polytechnic who availed themselves to be interviewed for this research and my best friend, Isaac Jonah Koomson and Miss Lucy Opoku Bediako who assisted me in data entry and type setting. I say God bless you all. iv University of Ghana http://ugspace.ug.edu.gh TABLE OF CONTENTS CONTENTS PAGES Declaration ................................................................................................................................ i Abstract ................................................................................................................................... ii Dedication ............................................................................................................................... iii Acknowledgement ................................................................................................................. iv Table of Contents .................................................................................................................... v List of Tables ........................................................................................................................ viii List of Figures ........................................................................................................................ ix List of Abbreviations ............................................................................................................... x CHAPTER ONE 1.0 Introduction ....................................................................................................................... 1 1.1 Purpose of the study .......................................................................................................... 5 1.2 Research Questions ............................................................................................................ 5 1.3 Data Collection Procedure ................................................................................................. 6 1.4 Outline of the Thesis ......................................................................................................... 8 CHAPTER TWO - LITERATURE REVIEW 2.0 Introduction ...................................................................................................................... 11 2.1 Evolution of Telecommunication .................................................................................... 11 2.2 Ghana‟s Telecommunication ........................................................................................... 16 CHAPTER THREE - REVIEW OF METHODS 3.0 Introduction .................................................................................................................... 19 v University of Ghana http://ugspace.ug.edu.gh 3.1 Factor Analysis ............................................................................................................... 19 3.1.1 Factor Analysis Model ................................................................................................. 20 3.2 The Principal Component Solution of Factor Model ...................................................... 23 3.3 Determining Suitable Sample Size .................................................................................. 25 3.4 Test for Internal Reliability (or consistency) ................................................................... 27 3.5 The Minimum Standard Test for Factor Analysis ........................................................... 27 3.5.1 Bartlett‟s Test of Sphericity .......................................................................................... 27 3.5.2 Kaiser-Meyer-Olkin (KMO) Test of Sampling Adequacy .......................................... 29 3.6 Number of Factors to Extract .......................................................................................... 30 3.6.1 Determination Based on Eigenvalues ........................................................................... 30 3.6.2 Determination Based on Scree Plot ............................................................................. 30 3.6.3 Significance Test of Eigenvalues ................................................................................. 31 3.7 Test of Goodness-of-Fit of the Factor Model ................................................................. 31 3.8 Rotation Factor Solution ................................................................................................. 34 3.8.1 Varimax Rotation ......................................................................................................... 36 3.9 Estimating Factor Score ................................................................................................... 37 3.9.1 Interpretation of Estimated Factor Score ...................................................................... 38 CHAPTER FOUR - ANALYSIS AND RESULTS OF THE RESEARCH 4.0 Introduction ...................................................................................................................... 40 A. Preliminary Analysis ..................................................................................................... 40 4.1 Distribution of Age and Gender ...................................................................................... 41 4.2 The Mean and Standard Deviation of the Mobile Phone Attributes ................................ 42 4.3 Usage of Mobile Phone Network .................................................................................... 43 4.4 Bar Charts of the Mobile Phone Attributes ..................................................................... 46 vi University of Ghana http://ugspace.ug.edu.gh 4.5 Correlation Matrix of the variables (mobile phone attributes) ....................................... 60 4.6 Test of Reliability (or consistency) .................................................................................. 62 4.7 Kaiser-Meyer-Olkin (KMO) Test and Bartlett‟s test ...................................................... 62 B. Further Analysis ............................................................................................................. 63 4.8 Number of Factors to Extract .......................................................................................... 63 4.8.1 Use of Eigenvalue Analysis ......................................................................................... 63 4.8.2 Scree Plot ..................................................................................................................... 65 4.9 Reproduce Correlation Matrix ........................................................................................ 66 4.10 Un-rotated Factor (Component) Matrix ......................................................................... 67 4.11 Varimax Rotated Factor (Component) Matrix ............................................................... 68 4.11.1 Factor Rotation ........................................................................................................... 69 4.11.2 Final Factor Solution................................................................................................... 71 4.12 Estimating Factor Score ................................................................................................. 71 CHAPTER FIVE - SUMMARY, CONCLUSION AND RECOMMENDATIONS 5.0 Introduction ...................................................................................................................... 73 5.1 Summary of Findings ....................................................................................................... 73 5.2 Discussion of Findings ..................................................................................................... 74 5.3 Conclusion and Recommendations ................................................................................. 77 References ............................................................................................................................. 79 Appendices ............................................................................................................................ 83 vii University of Ghana http://ugspace.ug.edu.gh LIST OF TABLES Table 4.1: Distribution of Age and Gender of the Respondents ............................................ 41 Table 4.2: Mean and Standard Deviation of the Mobile Phone Attributes ........................... 42 Table 4.3: Usage of Mobile Phone Network ........................................................................ 43 Table 4.4: Usage of Mobile Phone Network by respondents in Percentage ......................... 43 Table 4.5: Correlation Matrix ............................................................................................... 60 Table 4.6: Test of Internal Reliability (Consistency) ............................................................. 62 Table 4.7: Kaiser-Meyer-Olkin (KMO) and Bartlett‟s Test ................................................. 62 Table 4.8: Total Variance Explained ..................................................................................... 64 Table 4.9: Reproduce Correlation Matrix .............................................................................. 66 Table 4.10: Un-rotated Component Matrix ........................................................................... 67 Table 4.11: Varimax Rotated Component Matrix ................................................................. 68 viii University of Ghana http://ugspace.ug.edu.gh LIST OF FIGURES Figure 1: Distribution of Wider Coverage of Mobile Phone Network ................................. 46 Figure 2: Distribution of Reliable Network .......................................................................... 47 Figure 3: Distribution of lower tariff of a Network ............................................................... 48 Figure 4: Distribution of Frequent Promotions and Motivation Packages of Mobile Phone Network ..................................................................................................................... 49 Figure 5: Distribution of Close Relations (eg. Friends, Family) who use same Network .... 50 Figure 6: Distribution of Fast Internet Speed ........................................................................ 51 Figure 7: Distribution of Cheaper Starter Pack ..................................................................... 52 Figure 8: Distribution of Strong Area Network .................................................................... 53 Figure 9: Distribution of Mobile Phone Banking Services .................................................... 54 Figure 10: Distribution of Games of Chance of Mobile Phone Network ............................. 55 Figure 11: Distribution of the Network has Good Advert .................................................... 56 Figure 12: Distribution of Regular Sponsorship of National Events ..................................... 57 Figure 13: Distribution of Charity Work of the Network Operators ..................................... 58 Figure 14: Distribution of Long Time Usage of a Network .................................................. 59 Figure 15: Scree Plot of Eigenvalues and Components ........................................................ 65 ix University of Ghana http://ugspace.ug.edu.gh LIST OF ABBREVIATIONS AT&T - American Telephone and Telegraph ATM - Automated Teller Machine C.E.Os - Chief Executive Officers HSPA - High Speed Packet Access I.C.T - Information Communication Technology IP - Internet Protocol KMO - Kaiser-Meyer-Olkin MTN - Mobile Telecommunication Network NCA - National Communication Authority SPSS - Statistical Package for Social Science TAT-1 - First Transatlantic Telegraph Cable VoIP - Voice over Internet Protocol Wi-Fi - Wireless Fidelit x University of Ghana http://ugspace.ug.edu.gh CHAPTER ONE 1.0 INTRODUCTION Telecommunication (communication by telephone) is the transmission of information from one point to another by electrical and electronic means. It can also be defined as any transmission or reception of signs, signals, writings, images, sounds and other forms in which intelligence can be presented by means other than human transport and presentation. This means telecommunication has taken over other means of communication- face to face delivery of information and letter writing. Early Telecommunication began with the use of smoke signals and drums. Talking drums were used by natives in Africa, New Guinea and South America, and smoke signals in the North America and China. These systems were often used to do more than merely to announce the presence of a military camp (Baker & Burton, 2000). st The 21 century has seen massive development in telephony. Internet Protocol (IP) telephony also known as Internet Telephony or Voice over Internet Protocol (VoIP), is a disruptive technology that is rapidly gaining grounds against traditional telephone network. These new developments in telecommunication seek to address the communication problems of businesses and Government operations as well as individuals. In Ghana, it is believed that total number of mobile phones is more than the human population, this is due to multiple subscription (Daily Graphic, November 15, 2013). This suggests that telephone communication has become one of the necessities of life after food, shelter, and clothing. 1 University of Ghana http://ugspace.ug.edu.gh It is estimated that over 6.9 billion customers worldwide use mobile phones and in Ghana, mobile phone subscribers amounted to 26,336,000 distributed among the six mobile phone network providers - MTN, Tigo, Vodafone, Airtel, Expresso and Glo (Akakpo, 2008). This telecommunication industry covers an immense assortment of technologies that send information over long distances. The sector, relate to businesses which provide these technical services. The mobile phone, satellites technology, the internet and telephony, are at the centre of the telecommunications sector. The sector now covers: 1. Mobile phone operators (i.e. Vodafone, MTN, Tigo, Expresso, Airtel and Glo) 2. Message communication services (e–mail, Facebook, YouTube and Twitter) 3. Distributors of cable and pay television (e.g. virgin media, sky) 4. Manufacturers of accessories such as earphones, adaptors, cables, and Bluetooth products and accessories. The competition in telecommunication industry (mobile phone network provider) is now very keen. In Ghana, all the providers are putting in new strategies in order to have the largest market share. Some of these strategies or innovations are: 1. Increasing their coverage area. 2. Making their network reliable and affordable. 3. Making frequent promotions. 4. Providing banking services through mobile phones/telephony. 5. Organizing game of chance (lottery) Studies have revealed that social activities and the travel demand, will be influenced by the use of new Information and Communication Technologies (I.C.Ts) such as the Internet. 2 University of Ghana http://ugspace.ug.edu.gh These I.C.Ts offer new ways of communication, these alternatives for face-to-face communication have raised speculations about the consequence of I.C.Ts for social interactions (Miller, 1980). Price Waterhouse Coopers (2009) report from 23 countries revealed that disruptive change is a constant feature of the communication industry and results from this survey showed that communication C.E.Os see little sign of the pace and scale of change diminishing in the future. It is surprising that 36% of communication C.E.Os is planning to make fundamental strategic changes. They were also anxious about the security of their supply chains and a significant proportion (19%) believe that lack of basic infrastructure in some markets is likely to be a serious problem for their business. The report also revealed that communication C.E.Os who are not at all confident of being able to deliver growth is noticeably higher (14%). Research indicates that the implications of the internet and the mobile phone are complex and dependent on the type of activity, persons involved, technologies and socio–physical context in which they are embedded (Miller, 1980). Choo and Mokhtarian (2006) in their study of Telecommunications and travel demand and supply showed an empirical result which strongly support the hypothesis that telecommunications and travel are complementary. That is, as telecommunication demand increases, travel demand increases, and vice versa. Ren and Kwan (2008) found that the impact of internet activities on people‟s activity-travel patterns are significantly different across gender. In general, internet use for maintenance 3 University of Ghana http://ugspace.ug.edu.gh purposes has a greater impact on women‟s activity-travel in the physical world, while internet use for leisure purposes affects men‟s physical activities and travel to a greater extent. The effort being made by the six mobile phone operators to make telecommunication (telephone communication) affordable, reliable and faster is an indication that the companies are aware of the other needs of the society in general than just making and receiving a call. These are the expectations of the mobile phone subscribers. The mobile phone operators- MTN, Vodafone, Tigo, Airtel, Glo and Expresso in the Ghanaian market have the potential of throwing the consuming public in a maze of choice making. It is believed that mobile phone network subscribers consider some underlying (latent) factors before making a choice. These underlying (latent or hidden) factors can be determined by a multivariate statistical technique known as factor analysis. Hence the topic, “Factor Analysis of Customer’s Preference for Mobile Phone Network’’. Factor analysis is a statistical technique used to describe the covariance relationships among many variables (mobile phone attributes) in terms of a few underlying or unobserved random quantities called factors (Rommel, 1970). In this research, factor analysis is performed using principal component (pc) method of factoring. Principal component (pc) is a statistical procedure used to form new variables (underlying or latent factors) which are linear combinations of original variables (mobile phone network attributes). In principal component method of factoring the maximum number of unobserved or new variables (latent or underlying factors) that can be formed is less or equal to the number of observed or original variables (mobile phone attributes). 4 University of Ghana http://ugspace.ug.edu.gh 1.1 PURPOSE OF THE STUDY The purpose of this study is to: 1. Compile the ratings of some attributes of some mobile phone network providers by some sections of Cape Coast Polytechnic population. 2. Identify the correlations that exist between these attributes of mobile phone network. 3. Use factor analysis to evaluate these ratings in order to determine the main underlying (latent) factors that influence one‟s (subscriber‟s) choice of mobile phone network. 1.2 RESEARCH QUESTIONS In this research, customer‟s preference for mobile phone network may be defined as the conscious effort to make certain expenditure choices based on convictions about a particular network. Such personal convictions may be influenced by a number of conditions such as: 1. Are the six mobile phone operators on the market equally good? 2. Does the selected network have the desired qualities (factors) because of the rapid change over from one network to other? 3. Does the network operator provide the customers telecommunication needs and expectation? 4. Do the subscribers want the kind of reward schemes that the network operators provide? 5. Do the network operators provide the type of products and services that the subscribers expect? (eg. Mobile TV, ability to use phone for monetary transactions). These sources of influence on choice of mobile phone network raises a major question which must be answered; “what major (latent or underlying) factors ultimately influence the choice of a mobile phone network‟‟. 5 University of Ghana http://ugspace.ug.edu.gh 1.3 DATA COLLECTION PROCEDURE To achieve the objectives set out in the previous sections a survey of 500 respondents comprising students, lecturers and administrative staffs were selected by use of proportional allocation. This was done to ensure homogeneity within stratum, so that any group of members selected from the stratum is a good representation of the entire population. It is expected that the results of the analysis of opinions from these samples (strata) on almost all the variables (mobile phone attributes) defined below in the study would show the attitude of mobile phone network subscribers. In this study, the respondents were asked to indicate their opinion on fourteen attributes of their stated most used mobile phone network–MTN, Vodafone, Tigo, Airtel, Expresso and Glo. The selected attributes of mobile phone network are: X1........................................wider coverage X2........................................network reliable X3........................................lower tariff X4........................................frequent promotion/rewarding/motivations X5........................................close relations (eg. friends or family) use same network X6........................................fast internet speed X7........................................cheaper starter pack (chip) X8........................................strong network in one‟s area X9........................................mobile phone banking services X10.......................................frequently organize games of chance X11........................................has good advert X12........................................regular sponsorship of national events (eg. football) X13........................................encouraging charity work 6 University of Ghana http://ugspace.ug.edu.gh X14........................................used for a long time In this research, respondents were asked to indicate the level of importance attached to each of the following indicators by use of questionnaire. The following five point‟s rating scale was used: 1. Strongly disagree 2. Disagree 3. Undecided 4. Agree 5. Strongly agree In the questionnaire, a respondent indicating 4 or 5 against a variable suggests that the person really pays much attention to that attribute (variable), 1 or 2 indicates that the person does not attach much importance to the attributes and finally 3 indicates that the attributes in question at times enjoy some level of importance when choosing a mobile phone network. Factor analysis (the main tool used in this research) is a statistical technique that is used for data reduction and summarization. In this procedure a large number of variables (mobile phone attributes) most of which are correlated is reduced to a manageable level. Factor analysis attempts to explain the correlation between the observations in terms of the underlying (latent) factors which are not directly observable. Factor analysis closely resembles principal components analysis (Cattell, 1966). Both techniques use linear combinations of variables to explain sets of observations on many variables. The combination of variables is a tool for simplifying the interpretation of the observed variables. In factor analysis, the intrinsic interest is in the “underlying factors” or “hidden factors” 7 University of Ghana http://ugspace.ug.edu.gh (latent or unobservable variables called factors). Linear combinations are formed to derive the latent or “hidden factors”. The observed variables (mobile phone attributes) are relatively of little interest. The objective in factor analysis method is to identify few factors out of the lot that seek to explain the variation in the original data set. These latent factors would be responsible in explaining the attitude of the subscribers in the population when it comes to the selection of a mobile phone network. Cape Coast Polytechnic, used as a case study, is a tertiary institution in Cape Coast, the capital of the Central Region of Ghana. The institution is located at north–eastern part of the metropolis. It has a total population of N  3284 made up of students N1  2884 , teaching staff N2 102 and administrative staff N3  298 . Cape Coast Polytechnic is strategic because in each stratum (group) every unit uses mobile phone. 1.4 OUTLINE OF THE THESIS In this section, we outline the content within each of the five chapters of the thesis. The first chapter of this thesis is the introduction which covers the background to the study, its objective, research questions, sources of data and data collection procedures and design. In the background study, the history of stages of innovational efforts made by the scientists to make telephone communication faster and at every place of human habitat to satisfy the consuming public is given. It is then followed by the statement of the purpose of the study. This subsection attempts to justify and explain the need for the choice of the research topic. 8 University of Ghana http://ugspace.ug.edu.gh Moreover, the statistical tool (factor analysis) that was used to analyze the data is explained. Next on the list in the introduction is the statement of the research questions, sources of data and data collection procedure. It considers how data was generated and its scope. This sub- section discusses the choice of the study site and states the conditions under which findings of the study are most relevant. Outline of the thesis is the last section under introduction and it gives a brief overview of the content of the study. Chapter two covers literature review. It discusses the studies and discoveries made so far in telecommunication and the findings that have emerged. The history in telecommunication is also considered. Chapter three reviews the important methods used in the analysis of the data. The main method considered is factor analysis. Relevant exploratory tools are used to obtain an idea of the general pattern of the raw data. Chapter four comprises of two sections, preliminary and further analysis. In exploring the data, relevant pictorial representation (bar graph) and basic summary statistics - mean, standard deviation and correlation matrix (Preliminary analysis) were used. In this same chapter, factor analysis (Further analysis) is used to extract the underlying (latent) factors that influence the subscribers‟ choice of mobile phone network. Further steps are taken to ensure that the extracted factors give the most appropriate factor solution that could be obtained for the data. These measures check the goodness-of-fit of the model obtained and the interpretation given to the factors (attributes) are most appropriates. In Chapter five, the results of the preliminary and further analysis are discussed. The challenges that were encountered in the research process are also mentioned and discussed. 9 University of Ghana http://ugspace.ug.edu.gh Conclusions of the analysis are reached whereby relevant recommendations as well as further research areas in telephone communication are mentioned. 10 University of Ghana http://ugspace.ug.edu.gh CHAPTER TWO LITERATURE REVIEW 2.0 INTRODUCTION The benefit of telephone communication as compared to the face-to-face delivery of messages cannot be over emphasized – it saves time, removes travelling risk, is very fast and less expensive. 2.1 EVOLUTION OF TELECOMMUNICATION Early telecommunication began with the use of smoke signals in North America and parts of Asia (eg. China) and talking drums by natives in Africa, New Guinea and South America. These systems were often used to do more than merely to announce the presence of a military camp (Baker & Burton, 2000). In 1792, a French engineer, Claude Chappe built the first visual telegraphy (Semaphore) system between Lille and Paris. This was followed by a line from Strasbourg to Paris. In 1794, a Swedish engineer, Abraham Edelcrantz built a quite different system from Stockholm to Drottingholm. As opposed to Chappe‟s system which involved pulleys rotating beams of wood, Edelcrantz system relied only upon shutters and was therefore faster. However, Semaphores as a communication system suffered from the need for skilled operators and expensive towers often at intervals of only ten to thirty kilometres. As a result, the last commercial line was abandoned in 1880, see (Arthur, 1906). A very early experiment in electrical telegraphy was an “electrochemical” telegraphy created by a German inventor Samuel Thomas Von Sommerring in 1809, based on an earlier less 11 University of Ghana http://ugspace.ug.edu.gh robust design of 1804 by Spanish polymath and Scientist Francisco Salva Campillo (Jones and Sommerings, 1965). Their design used multiple wires (up to 35) in order to visually represent almost all Latin letters and numerals. Thus messages could be conveyed electrically up to a few kilometres with each of the telegraph receiver‟s wires immersed in a separate glass tube of acid. An electrical current was sequentially applied by the sender through the various wires representing each digit of message at the recipient end. The current electrolyzed the acid in the tubes in sequence, releasing streams of hydrogen bubbles next to each associated letter or numeral. The telegraph receiver‟s operator would visually observe the bubbles and then record the transmitted message (Jones & Sommerings, 1965). The first successful transatlantic telegraph cable (TAT-1) was completed on July 27, 1866, allowing transatlantic telecommunication for the first time, this international use of the telegraph has sometimes been dubbed the “Victorian Internet” (Dibner, 1959). The first commercial telephone services were set up in 1874 and 1879 on both sides of the Atlantic in the cities of New Haven and London (Arthur, 1906). Alexander Graham Bell held the master patent for telephone that was needed for such services in both countries. The technology grew quickly from this point with inter-city lines being built and telephone exchanges in every major city of the United States by the mid-1880s (Coe and Lewis, 1995). For a short period of time acoustic telephones were marketed commercially as a niche competitor to the electrical telephone, as they preceded the latter‟s invention and didn‟t fall within the scope of its patent protection. When Alexander Graham began competing for customers, acoustic telephone makers quickly went out of business (Kolger & Jon, 1986). 12 University of Ghana http://ugspace.ug.edu.gh th During the second half of the 19 century inventors tried to find ways of sending multiple telegraph messages simultaneously over a single telegraph wire by using different modulated audio frequencies for each message. These inventors included Charles Bourseul, Thomas Edison, Elisha Gray, and Alexander Graham Bell. Their efforts to develop acoustic telegraphy in order to significantly reduce the cost of telegraph messages led directly to the invention of the telephone called the „Speaking telegraph‟ (Daniel & McVeigh, 2013). Credit for the invention of electric telephone is frequently disputed and new controversies over the issue have arisen from time-to-time. Charles Boursel, Antonio Meucci, Johann Philip Resis, Alexander Graham Bell and Elisha Gray, amongst others, have all been credited with telephone‟s invention (Coe & Lewis, 1995). Evenson and Edward (2000) considered the question of whether Bell and Gray invented the telephone independently, and if not whether Bell stole the invention from Gray. This controversy is narrower than the broader question of who deserves credit for inventing the telephone, for which there are several claimants (Evenson & Edward, 2000). The first commercial telephone exchange in the world was opened at New Haven, Connecticut, with 21 subscribers on January 28, 1878 (Huurdeman & Anton, 2003). A telephone exchange is a telephone system located at service centres, central offices responsible for a small geographical area that provided the switching or interconnection of two or more individual subscriber lines for calls made between them, rather than requiring direct lines between subscriber stations. This made it possible for subscribers to call each other at homes, business and public places. These made telephony an available and 13 University of Ghana http://ugspace.ug.edu.gh comfortable communication tool for everyday use, and it gave the impetus for the creation of a whole new industrial sector. The history of mobile to two-way radio permanently installed in vehicles such as taxi cabs, police cruisers, railroad trains, and the like. Later versions such as the so called transportable or „bag phones‟ were equipped with a cigarette lighter plug so that they could also be carried, and thus could be used as either mobile two-way radio or as portable phones by being patched into the telephone network. In December 1947, Bell laboratory engineers Douglas H. Ring and W. Rae Young proposed hexagonal cell transmissions for mobile phones (Richard & John, 2010), (Wheen & Andrew, 2011). The technology did not exist then and the radio frequencies had not yet been allocated. Cellular technology was undeveloped until the 1960s, when Richard H. Frenkiel and Dr. Joel Engel of Bell Laboratory developed the electronic (Wheen & Andrew, 2011). On April 3, 1973 motorola manager Martin Cooper placed a cellular phone call (in front of reporters) to Dr. Joel Engel, head of research at American Telephone and Telegraph (AT & T‟s) Bell Laboratory (Wheen & Andrew, 2011). This began the era of the handheld cellular mobile phone. Meanwhile the 1956 inauguration of TAT-1 cable and later international direct dialing were important steps in knitting together the various continental telephone networks into a global network. Internet Protocol (I.P) which uses a broadband internet service to transmit conversations as data packets also competes with mobile phone networks offering free or lower cost service via Wireless Fidelity (Wi-Fi) hotspots. Modern telecommunication also used Voice over 14 University of Ghana http://ugspace.ug.edu.gh Internet Protocol (VoIP) which is used on private wireless networks which may or may not have a connection to the outside telephone network. Globally, mobile subscriptions (including multiple subscriptions) are expected to reach 6.9 billion in 2013. The top mobile operator worldwide in terms of connections is China mobile, followed by Vodafone group (Akakpo, 2008). It has been projected that, 2013 and 2014 will see a positive growth in mobile phone infrastructure expenditure as carriers are forced to improve and deploy new network to cope with demand. Also by 2017 around 45% of mobile traffic is expected to be offloaded from Wireless Fidelity (Wi-Fi) (Akakpo, 2008). Moreover, the cost of acquiring a customer has grown along with the increase in smart phone uptake. Subsidizing handset is an expensive exercise and it has become even more important for telecommunication companies to retain the customer once they are on board. In addition, lowering roaming charges also encourages goodwill at both a regulatory and consumer level and lessons the chance of bill-shock. To improve the customer experience, it is essential that the service and available data information is of the highest possible quality and real time processing development can assist with this. There is currently a lack of high customer expectation in telecommunication market as a whole – and much can be done to improve this situation. Price Waterhouse Coopers (2009) reported that, many communication‟s CEOs are reconsidering how best to manage innovation, they are repositioning their portfolios to focus on developing new products and services and fine tuning existing products and services. But 60% intend to adopt new business models in response to a fast-changing environment. 75% of CEOs think talking to the customers who buy their product and services would be the best 15 University of Ghana http://ugspace.ug.edu.gh of their time. Predictably, perhaps, communication CEOs are pinning their hopes for future growth on emerging markets rather than the developed market-as indeed are their peers in other sectors. And while most CEOs with plans to expand abroad are focusing on China, 26% of communication CEOs prefer Brazil since they believe it will be a key growth market. 2.2 GHANA’S TELECOMUNICATION Telecommunication is one of the main economic sector of Ghana, due to the Ghana liberal policy around Information and Communications Technology (ICT) (World Bank, 2013). World Bank (2013) reported that, the main sector of investment in Ghana are 65% is for ICT, 8% for communication and 27% is divided for public administration. Ghana‟s telecommunication statistics indicated that as of 2012, there were 284,981 telephone lines (landlines) in operation and as of 2013, there were 26,336,000 cell phone lines in operation which is more than the country‟s population at the time, this is due to multiple subscriptions. Since launching the first cellular mobile network in Sub-Saharan Africa in 1992, Ghana has become one of the continent‟s most vibrant mobile markets with now six (6) competing mobile phone operators including regional heavy weights such as Mobile Telecommunication Network (MTN), Vodafone, Millicom (Tigo), Bharti Airtel (formerly, Zain) and Expresso (formerly, Kasapa). The entry of Nigeria‟s Globacom (Glo) as the sixth player in 2012 delivered another boost to the telecommunication sector. Ghana has one of the most competitive telecommunication markets in the sub region and was a pioneer in developing mobile telephony and data service. It was also among the first on the continent to connect to the internet (Akakpo, 2008). 16 University of Ghana http://ugspace.ug.edu.gh According to National Communication Authority‟s (NCA‟s) homepage (2012), the market share of the service providers are MTN (45.8%), Vodafone (20.53%), Tigo (14.44%), Airtel (12.46%), Globacom (6.12%) and Expresso (0.65%). Ansah et al. (2013) in their study of prediction of subscribers‟ brand switching behaviour and ergodic market share of network service providers in Ghana reveals that the three most preferred operators are MTN (64.9%), Tigo (38.0%) and Vodafone (37.7%). Low user penetration in the early part of the century was largely due to the high cost of service coupled with unreliable networks and a poor quality of service. In recent years lower pricing has filtered down well to consumers with Ghana Telecom being one of several operators which have invested in national networks to extend broadband availability deeper into rural areas (Akakpo, 2008). The mobile market is well served by six competing players. Services based on High Speed Packet Access (HSPA) technology have helped extend broadband availability. This has improved the growth potential of m-commerce and m- banking services. MTN Ghana‟s mobile money service is very popular, complimented by its new “ATM cash out” services. The launch of mobile number portability in mid-2011 has also been a catalyst for competition between players with the number of portings by mid-2013 having increased by 21% year-on-year. As at 2013 the telecommunication penetration by services in Ghana stands at 17% for internet, 112% for mobile telephony and 1.2% for fixed- line telephony. This indicates that competition among mobile phone companies in Ghana is an important part of the telecommunications industry growth of Ghana with companies obtaining more than 80 per 100 persons as mobile phone and fixed phone users (Akakpo, 2008). 17 University of Ghana http://ugspace.ug.edu.gh On November 26, 2013, Daily Graphic reported an introduction of a new service, „Triple play‟ which provides telephony, internet and television in a single box into the Ghanaian market and subsequently to other African countries. The service “Triple play” which will be introduced by K3 Telecom AG (a Swiss telecommunication company) will make it possible for consumers to have these three services (Telephony, Internet, Television) at a go, once they have purchased the box. This suggests that in the years ahead, consumers must expect new technological advancement in the telecommunication industry. 18 University of Ghana http://ugspace.ug.edu.gh CHAPTER THREE REVIEW OF METHODS 3.0 INTRODUCTION This chapter reviews the theories and methods that relate to the main analysis of the data. Data in this research are multivariate in nature because the variables (mobile phone attributes) considered are more than two. In this research factor analysis as a multivariate statistical technique was used to analyze the data. 3.1 FACTOR ANALYSIS Factor analysis is a statistical technique that is used for data reduction and summarization. It is used to describe, if possible the covariance relationships among many variables (mobile phone attributes) in terms of a few underlying, but unobserved random quantities called factors. The fourteen (14) mobile phone attributes (observed variables) will be reduced to smaller number of unobserved variables called factors. In this research, the fourteen mobile phone attributes (variables) most of which are correlated is reduced to manageable level using factor analysis. In many scientific fields, particularly behavioural and social sciences, variables such as “intelligence” or “leadership quality” cannot be measured directly. Such variables, called “latent” variables, can be measured by other “quantifiable” variables, which reflect the underlying variables of interest. Factor analysis used in this study, attempts to explain the correlation between the observations (attributes) in terms of the underlying (latent) factors which are not directly observable. This statistical technique (factor analysis) was originally developed to explain 19 University of Ghana http://ugspace.ug.edu.gh student performance in various courses and to understand the link between grades and intelligence. Spearman (1909) hypothesized that a student performance in various courses are inter- correlated and their inter-correlations could be explained by a single latent factor of student‟s general intellectual ability and a second set of factors reflecting the unique qualities of the individual courses. Factor analysis closely resembles principal component analysis. Both techniques use linear combinations of variables (attributes) to explain sets of observations on many variables. Thus, factor analysis is carried out in this research using principal component method of factoring. The combination of variables (underlying factors) is a tool for simplifying the interpretation of the observed variables (mobile phone attributes). The linear combinations of the observed variables (mobile phone attributes) are formed to derive the underlying (latent) factors. In factor analysis, the main interest is in the “factors” or “latent factors” of the observed variables (mobile phone attributes). The observed variables are relatively of little interest. 3.1.1 Factor analysis model In orthogonal factor model, the observable (mobile phone attributes) random vector X with p components has mean  and covariance matrix Σ . The factor model postulates that X is linearly dependent upon a few unobservable (latent) random variables f1, f2 ,..., fm called common factors and p additional sources of variation 1,2 ,..., p called errors or specific factors. 20 University of Ghana http://ugspace.ug.edu.gh In particular, the factor analysis model is given as: X1 1  l11 f1  l12 f2  ... l1m fm 1 X2 2  l21 f1  l22 f2  ... l2m fm 2 . . . . . . . . . X   l p p p1 f1  lp2 f2  ... lpm fm  p where p  number of mobile phone attributes. m  1 ,2, factors, m  p l = factor loadings (correlation between the factors and mobile phone attributes) The above expression can be represented in matrix form as: X – μ = LF +ε (3.1) X = LF +μ +ε where μ  ( p1) mean of variables (attributes) X  ( p1) vector of the mobile phone attributes (indicator variables) ε  ( p1) specific variance F  (m1)common factors L  ( pm) matrix of the factor loadings. The unobservable random vectors F and ε are independent whereas these assumptions must be satisfied. E F  0 , CovF  I E(ε)  0, Cov (ε)  Ψ where, Ψ is a diagonal matrix. 21 University of Ghana http://ugspace.ug.edu.gh The orthogonal factor model of a covariance structure for X from equation (3.1) is: (Xμ)(Xμ)  (LFε)(LFε)  (LFε)(LF)  (LFε)ε  LF LF  ε LF  LFε  εε (3.2) So that, Σ Cov X  E(Xμ)Xμ , taking the expectation of equation (3.2) LE FFL E(εF)LLE FεE(εε) LL Ψ (3.3) where, L   p pmatrix of factor loadings and Ψ  ( p p) matrix of specific variance(errors). Thus, the covariance structure for the orthogonal model is: (i) C ov(X) LL+ Ψ or Var(x )  l 2 i i1  l 2 i2  ... l 2 im  (ii) C ov(xi xk )  li1lk1  ... limlkm (iii) C ov(X,F) = L or C ov(X iFj )  lij The model X – μ  LFε is linear in the common factors. If the p responses of X mobile phone attributes are in fact, related to underlying factors but the relationship is non-linear such as X1 1  l11 f1 f2 1 , X2 2  l21 f2 f3 2 and so on, then the covariance structure, CovX LLΨ may not be adequate. aij i The factor loading lij is estimated as: lij  , where aij  weight of xi , si 22 University of Ghana http://ugspace.ug.edu.gh si  the standard deviation of xi and i = the eigenvalue of xi . p p   S 2 2i 1  S2  ... S 2 p i  trace(S) i1 i1 S 2 is the sample variance of xi whereas the eigenvectors of the mobile phone attributes are i estimated by equation, (I)e  0  ee 1 Thus, the first eigenvector e1 corresponding to the first eigenvalue 1 is obtained as (1I)e1  0 (3.4)  e1e1 1 multiplying equation (3.4) by e1 gives e( I)e  0  e1 e1  1 1 1 1 , I  identity matrix The sample variance, var(x )  s2 due to the specific factors is often called the uniqueness or i i m specific variance  i and it is given as:  i  s 2 h2 or i i  i 1 l im i1 Communality h2 is the amount of variance a mobile phone attribute shares with all other i mobile phone attributes under consideration or the proportion of variance explained by the common factors. The communality denoted by h2 is given by: i m 2 hi  l 2 2 i1  li2  ... l 2 im l 2 ij , j1 where i  1,2,...,14 number of mobile phone attributesand m  number of factors, m  i  p One can notice that the communality h2 is the sum of squares of the factor loadings of the i ith mobile phone attribute on the common factors, and however, it is when m is small relative to i  p (14 mobile phone attributes) that factor analysis is most useful. 23 University of Ghana http://ugspace.ug.edu.gh 3.2 THE PRINCIPAL COMPONENT SOLUTION OF FACTOR MODEL Principal co mponent (pc) is a statistical procedure used to form new variables (underlying factors) which are linear com binations of original variables (mobile phone attributes). The principal component (pc) method of factor analysis of the sample covariance matrix S is ^ ^ ^ ^ ^ ^ specified in terms of its eigenvalue-eigenvector pairs (1,e1), (2 ,e2),..., ( p ,ep ).of the ^ ^ ^ mobile phone attributes, where   2  ...  p . 1 Let m  p be the number of common factors. The matrix of estimated factor (coefficients) loadings l is given as: ij ^ ^ ^ ^ ^ ^ L  ( 1 e1,  2 e2 ,...,  p e p ) (3.5) The estimated specific variances ψ of the mobile phone attributes are provided by the ^ ^ diagonal elements of the matrix S - LL , so that 11 0 . . . 0    0 22 0 . . 0   . 0 . .     . . .   . . .     0 0 . . .  pp  14 with   s  l2 (3.6) i ii im i1 m m 2 and communalities are estimated as: h2  l2 i i1  l 2  ... l2 l2 but l 1i2 im ij  ij j1 j1 and  1h2i i The principal component method of factor analysis of the sample correlation matrix R of the mobile phone attributes is obtained by starting with R in place of sample covariance matrix 24 University of Ghana http://ugspace.ug.edu.gh S . In princ ipal co mpo nent solution , the estimat ed factor loadings (coefficien ts) for a given ^ ^ ^ factor do not change as the number of factors increases. For instance, if m 1, L   1 e1 ^ ^ ^ ^ ^ ^ ^ ^ ^ and if m  2, L  ( 1 e1, 2 e ) where (1,e1) and ( 2 ,e2 ) are the first two eigenvalue-2 eigenvector pairs for sample covariance matrix S or sample correlation matrix R . It can be seen that, by the definition of Ψ the diagonal elements of S are equal to the diagonal ^ ^ elements of LL+ Ψ .However the off-diagonal elements of S are not usually reproduced by ^ ^ LL+ Ψ . The number of common factors m may be determined by a priori considerations, such as by theory or the work of other researchers, the choice of m can be based on the estimated eigenvalues in much the same manner as with principal components. 3.3 DETERMINING SUITABLE SAMPLE SIZE ( no ) Cochran (1977) proposed sample size determination formula given by: z2 p(1 p) n  2 , d 2 where n  sample size to be determined p  maximum possible population proportion d  acceptable margin of error 5% Z  value for selected  level of 5% 1.96 2 Krejcie and Morgan (1970) recommended that researchers should use p  0.5 as an estimate of maximum possible population proportion if the sample proportion cannot be obtained (is (1.96)2 (0.5)(0.5) unknown) from a previous comparable survey. Thus, n   384.16 (0.05)2 25 University of Ghana http://ugspace.ug.edu.gh Since the sample size exceeds 5% of the population 3284  0.05 164.2 , Cochran‟s (1977) correction formula is used to determine the final sample size, n 384 n    343.78 where, N  3284 , n  384 and n = final sample size to be 0 n 384 o 1 1 N 3284 determined. Considering a response rate of 70%, a minimum drawn sample size of 343.780.70  491.42 is required. Thus, sample of 500 respondents were taken from the population (students, administrative staff and teaching staff). Since the respondents are in three strata (students, teaching staff and administrative staff) the selection of response unit in nNh each stratum was determined by applying proportional allocation given by: nh  , N where n  sample size , N  total population , Nh  population in stratum h , nh  sample size from stratum h (students, teaching staff and administrative staff and h  stratum number . Thus, I represented n1  selected sample size from students , n2  selected sample size from teaching staff and n3  selected sample from administrative staff so that n1  n2  n3  n and N1  N2  N3  N , where n is the total sample size n  500 and N total population 2884 102 2884102 298  3284 . Therefore, n1  500  439 , n2  500 16 and 3284 3284 298 n . 3  500  45 3284 Thus, 439 respondents were selected from the students, 16 were selected from the teaching staff and 45 were selected from the administrative staff. 26 University of Ghana http://ugspace.ug.edu.gh Non-probability sampling (a sampling technique that does not employ the rules of probability theory) - quota and convenient sampling methods were used to select 500 respondents to answer the questionnaire. 3.4 TEST OF INTERNAL RELIABILITY (OR CONSISTENCY) Cronbach‟s alpha (α) is used to determine the reliability (or consistency) of the collected data. Cronbach‟s alpha (or reliability coefficient) is a measure of internal consistency that is how closely related a set of indicators (mobile phone attributes) are as a group. It is considered as a measure of scale reliability. A reliability coefficient (Cronbach‟s alpha) of 0.70 or higher is considered “acceptable” (Nunnally, 1978). pr Using standardized Cronbach‟s alpha defined as s tan dardised  (1 ( p 1)r) p( p 1) where p  number of mobile phone attributes and r  non-redundant correlation 2 coefficients (that is the mean of upper or lower triangular correlation matrix). 3.5 THE MINIMUM STANDARD TESTS FOR FACTOR ANALYSIS Before using factor analysis, the following minimum standard tests were applied to ascertain, if factor analysis is appropriate for the collected data 3.5.1 Bartlet’s Test of Sphericity Bartlet‟s test of sphericity is a test statistic that was used in this research to examine the hypothesis that the variables (mobile phone attributes) are uncorrelated in the population. In other words, the population correlation matrix is an identity matrix; each variable correlates perfectly with itself ( r 1) but has no correlation ( r  0) with other variable. 27 University of Ghana http://ugspace.ug.edu.gh Thus, if the mobile phone attributes are correlated then factor analysis can be used for the data. The test hypothesis is; Ho: Mobile phone attributes are uncorrelated against the alternative hypothesis H1: Mobile phone attributes are correlated. The test statistic is a chi-square transformation of the correlation matrix given by, 2 2 χ = -2 [1 - ( 2p + p + 2)] In L (3.7) where n  sample size500 , p  number of attributes 14 , n  p  2 n   Geometric mean   2   i  m  number of factors, m  p and L  i   i1    Arithmetic mean  p i   1   p i   i1  i (i 1,2,...14) are the eigenvalues of the component factors. The test statistic has  2 1  distribution with  p 1 p  2 degrees of freedom. Thus, at significance level 2   2of when the calculated value is greater than the table  2 , 1 value leads to 2 p1 p2 the rejection of the null hypothesis Ho . The conclusion is that: 1. The mobile phone attributes have no constant variance. 2. There are correlations among the mobile phone attributes. On the other hand if Ho is not rejected, it means, the mobile phone attributes are not correlated or the population correlation matrix is a ( p p) unit matrix and factor analysis would not be appropriate for the data. It can be seen from equation (3.7) that the test statistic 28 University of Ghana http://ugspace.ug.edu.gh strongly depends on the sample sizes ,n . That is, if n is very large, the value of In L causes the test statistic to be large. This suggests that there exist correlations among the variables for large values of n . This makes results of the Bartlett‟s test, highly dependent on the sample size (Bartlett, 1954). 3.5.2 Kaiser-Meyer-Olkin (KMO) Test of Sampling Adequacy Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy is an index that is used in this research to examine if factor analysis is appropriate for the data. High values between 0.5 and 1.0 indicate that factor analysis is appropriate. In the KMO test, if the value (index) is below 0.5 , it implies that factor analysis may not be appropriate for the data. The KMO index compares the magnitudes of the observed correlation coefficients with magnitudes of the partial correlation coefficients. Small values of the Kaiser-Mayer Olkin (KMO) statistic indicate that the correlations between pairs of mobile phone attributes cannot be explained by other mobile phone attributes and that factor analysis may not be appropriate for the data. Kaiser and Rice (1974) suggested the following guide for the interpretation of the KMO measure of sampling adequacy since there is no statistical test. Kaiser’s guide for interpreting KMO (Index) measure KMO measure Recommendation ≥ 0.90 marvelous 0.80+ meritorious 0.70+ middling 0.60+ mediocre 0.50+ miserable ≤ 0.50 unacceptable 29 University of Ghana http://ugspace.ug.edu.gh It can be seen from the guide that high value of KMO is desired. They suggested that the overall KMO measure should be greater than 0.8 , however, a measure above 0.60 is accepted. 3.6 NUMBER OF FACTORS TO EXTRACT This is where the information contained in the original variables (mobile phone attributes) is summarized. Smaller number of factors (latent factors) was extracted using combination of scree plot and eigenvalue greater-than-one rule. In this method, the main aim is to look for fewer (underlying or latent) factors that could do well in explaining the maximum amount of variation in the data set. 3.6.1 Determination Based on Eigenvalues (i) An eigenvalue represents the amount of variance associated with the factors. Zwich and Velicer (1986) suggested eigenvalue greater-than-one rule that, factors with eigenvalue greater than one contribute more in explaining the variance in the original set of variables (mobile phone attributes). In this procedure factors with eigenvalues greater than 1.0 are retained; the other factors are not included in the factor model. 3.6.2 Determination Based on Scree Plot This method was used to confirm the determination of extracted factors by eigenvalue- greater-than-one rule. The scree plot proposed by Cattell (1966), which is also called the elbow rule is a line graph of the eigenvalues i against the number of factors fi in order of extraction. In this method the number corresponding to where the plot exhibits an elbow is taken to represent the number of factors to be extracted. Typically, the plot has a distinct break between the steep slope of factors, with large eigenvalues and a gradual trailing off 30 University of Ghana http://ugspace.ug.edu.gh associated with the rest of the factors. This gradual trailing off is referred to as the Scree. Experimental evidence indicates that the point at which the scree begins denotes the true number of factors to be extracted. 3.6.3 Significance Test of Eigenvalues. The statistical significance of the separate eigenvalues are determined, and only those factors that are statistically significant are retained using the relation: 1 m  2p 1 8p 1 (3.8) 2 where m  number of factors to be extracted and p  number of mobile phone attributes 14 The number of common factors cannot exceed the largest integer satisfying the equation (3.8) and should be greater or equal to one (Morrison, 1976). 3.7 TEST OF GOODNESS-OF-FIT OF THE FACTOR MODEL Goodness of fit was used in this research to explain if the factor model is statistically good. Supposing m factors were extracted out of p mobile phone attributes, and denoting the loading matrix by L whose dimension is of  pm . The product LL is a  p p reproduced matrix of the correlation matrix of the indicator m variables. The diagonal elements of LL are called communalities and given by  lij of the j1 mobile phone attribute xi . If the m factors adequately approximate the correlation matrix R , 31 University of Ghana http://ugspace.ug.edu.gh then the difference R - LL is a diagonal matrix. The m factor model is then said to be adequate. Denoting this diagonal matrix by ψ , then R - LL = ψ where the elements of ψ are the m specific variances  i which is equal to  i 1 lij . The hypothesis to be tested to j1 determine the adequacy of the model is: H : Σ = ψ + LLo , the population covariance matrix can be estimated using m factors. H1 : Σ ,is any  p p symmetric positive definite matrix. The test statistic suggested by Bartlett (1954) is ^ ^^ ψ+ LL  2  N 1 1 2 p  4m5 In (3.9)  2  S ^ ^ where  and L are the solutions of the maximum likelihood equations given as: ^ ^ ^ ^  l(L, )  l(L, )  0 and  0   li ij the function l(L, ) is the logarithm of the likelihood function obtained from the Wishart 1 (np1) 1 density given by f (S) C S 2  2 exp 1 ntr1 s . If the null hypothesis is true, 2 2 2 as n becomes large, the statistic becomes a  distribution with v  1 ( pm)  pm 2   number degrees of freedom, where p  number of rows or columns of any symmetric positive matrix and m  number of common factors. 32 University of Ghana http://ugspace.ug.edu.gh At significant level of  , the null hypothesis of exactly m common factors is rejected if  2cal   2 table value By invariance property of the estimated loadings and specific variances, the same value of the test statistics would be obtained from a factor solution in terms of the correlation matrix. ^ ^^ ψ+LL Lawley and Maxwell (1971) showed that the value of the determinant ratio In is S ^ (s 2 ^ij  ij ) approximately equal to   , where  is the covariance of ith attribute with  iji j i j j th factor and given by: ^ m  ij  lijlij , i  j j1 th This is the ij off diagonal element of the product matrix LL . Thus, the approximate test statistic is given by: 2 ^   sij  ij   2  N 1 16 (2 p  4m5)   (3.10)  i j i j The test statistic above has the same asymptotic chi-square distribution as in equation (3.10) ^ but (s ij  ij ) is the ij element, resij  is the residual matrix of the m factor model. Thus, the test statistic of equation (3.10) above can be written as: 2  1  res 2   N 1 (2 p  4m5)  ij  (3.11)  6  i ji j If the factor model well approximates the correlation matrix R , the residual elements resij  are very small. A very small value of the sum of the test statistic in equation (3.11) 2 2 corresponds to the decrease in value of  . Thus, we fail to reject H0 , if the  value is 33 University of Ghana http://ugspace.ug.edu.gh smaller than the tabulated value  2 ,v . Where  is the level of significance and v  1 ( pm) 2  pm 2  degrees of freedom. Suppose m  0 , that is, 0 – factor model then the null hypothesis is reduced to: Ho : Σ  ψ, the covariance matrix is diagonal, that means the mobile phone attributes are independent against alternative hypothesis H1 : Σ  ψ, the covariance matrix is not diagonal, that is the mobile phone attributes are not independent. In this case the test statistic suggested by Bartlett (1954) is given as:  2  21 1 (2p11)6n  InL (3.12) n S 2 n where L   R sp with 1 p( p1) degrees of freedom. 2 2Sij i1 Now, for a diagonal matrix, the determinant is the product of the diagonal elements. Thus, if R is the  p p unit matrix, then R = 1 , and ln L  0 . Therefore, the value of  2 is very small if R is approximately equal to the unit matrix. When equation (3.12),  2 is very small, then we cannot reject Ho and conclusion is that the cal variables are independent. Thus, factoring cannot be applied. One can notice that, test of goodness-of-fit of the factor model is similar to the test of sphericity. 3.8 ROTATION FACTOR SOLUTION The un-rotated factor matrix indicates the relationship (correlation) between the factors and the individual mobile phone attributes. In this research, the un-rotated factor matrix is difficult to interpret, because the factors correlate with many attributes. Through rotation 34 University of Ghana http://ugspace.ug.edu.gh (varimax rotation) the complex factor matrix is transformed into a simpler matrix that is easy to interpret. In rotating the factors, we would like each factor to have non-zero, or significant loading or coefficient for only some of the fourteen mobile phone attributes. Rotation does not affect the communalities and the percentage of total variance explained, but the percentage of variance accounted for by each factor changes. Analytically the transformation ^ of the factor matrix corresponds to a rigid rotation (or reflection) of the coordinate axes. If L is the  pm matrix of estimated factor loadings obtained by principal component method of factoring then, ^ ^ * L = LT (3.13) Notice that TT' = T'T  I is a ( pm) matrix of rotated loadings. However, the estimated covariance (or correlation) matrix remains unchanged, since ^^ ^ ^ LL+ ψ = LTTL+ ψ ^ ^ ^ = * *L L + ψ (3.14) where T is the transformation matrix given by:  cos sin  cos sin  T    clockwise rotation, or T    anticlockwise sin cos  sin cos  rotation, and  , the angle of rotation. The equation (3.14), above, indicates that the residual matrix estimated by ^^ ^^  * *Sn - LL - ψ = Sn - L L - ψ of the reproduced matrix, remains unchanged. Moreover, the specific variance  2i and the communalities h are not altered. The complex i factor matrix is rotated to achieve a simpler structure that is easy to interpret. The varimax rotation is used (in this research) to achieve a pattern of loadings such that each variable 35 University of Ghana http://ugspace.ug.edu.gh (mobile phone attribute) loads highly on a single factor and has small-to-moderate loadings on the remaining factors (Morrison, 1976; Harman, 1976). ^ ^ ^ The new rotated loadings is determined by, *L = LT , where * L  ( p2) , estimated ^ transformation matrix of the factor loadings and L  ( p2) , estimated matrix of the factor loadings. 3.8.1 Varimax rotation. The varimax rotation suggested by Kaiser (1958) is used in this research to obtain a simple structure of fac tor loadings that can easily be interpreted. The final rotated coefficients ^ ^ * l * (factor loadings or correlation) is estimated by l  ij^ and the coefficients are scaled by the hi square root of the communalities. The normal varimax procedure selects the orthogonal transformation T that makes  2 p    m p  l *2 ij   1  *4  i1  v    lij   (3.15) p j1 i1 P     as large as possible, where p is the fourteen (14) mobile phone attributes. On scaling the * rotated coefficients lij has the effect of giving variables with small communalities relatively more weight in the determination of simple structure. After the transformation T is determined, the loadings l* are multiplied by hi so that the original communalities are ij preserved. The varimax method of factor rotation has the following properties: 36 University of Ghana http://ugspace.ug.edu.gh 1. Kaiser‟s (1958) maximand in the equation (3.15) above has a combined effect of equal weighting and normalization of the loadings for each attribute (variable). 2. Kaiser‟s varimax rotation function involves fourth powers of the loadings. 3. By considering variances of squared loadings for the columns rather than the rows of the loading matrix, varimax approach might focus more on identifying possible indicators (attributes) per factor. 3.9 ESTIMATING AND INTERPRETING FACTOR SCORE In principal component (pc) method of factor analysis, factor score (a score for an individual on a factor) can be estimated (for each respondent) if necessary to be used instead of the original variables (mobile phone attributes) in follow-up analysis. Multiple regression method (refined method) of estimating factor score is used in this research. Mathematically, factor score is a linear combination of p mobile phone attributes (original variables). For instance, the factor score for individual i on a given factor j is represented as:     Fij =B1 Xi1+B2 Xi2 +...+Bp Xip (3.16)   where Fij  estimated factor score for factor j for individual i , Bp  the estimated factor score coefficient for variable (factor loading on p mobile phone attribute) and th X = the p mobile phone attribute (observed indicator) for individual i . ip   Equation (3.16) is represented in matrix form as F = XB , where  F  (nm) matrix m factor score for the n individuals  B  ( pm) matrix of estimated factor (loadings) score coefficients X = n p matrix of mobile phone attributes 37 University of Ghana http://ugspace.ug.edu.gh The factor score coefficient for the standardized indicator variables (mobile phone attributes) are obtained from the factor score matrix when principal component (pc) method of factor analysis is used. For standardized indicators (mobile phone attributes)   F  ZB (3.17) Equation (3.17) can be written as: 1  1   1  ZF  ZZB or L = R B (when R correlation matrix is used) where L  (ZF) , n n n 1 x   R  (ZZ)and Z  i , i 1,2,...,14 attributes. Thus, the factor score coefficient matrix i n   is given by  1B R L and estimated factor scores by     F  1ZR L  -1 or F  L R Z (3.18)  1 where F  m1 row vector of m estimated factor scores, R   p p inverse of the  correlation matrix between the p mobile phone attributes, * L   pm estimated pattern matrix of loadings of p mobile phone attributes on m factors (components) and Z  row vector of p standardized mobile phone attributes (Z-score of 14 mobile phone attributes). One of the simplest ways to estimate factor scores for each individual involves summing raw scores corresponding to all mobile phone attributes loading on a factor. If a mobile phone attribute yields a negative factor loading, the raw score of the item is subtracted rather than added in the computations because the item is negatively related to the factor (Comrey & Lee, 1992). 38 University of Ghana http://ugspace.ug.edu.gh 3.9.1 Interpretation of Estimated Factor Score Factor score is the score of each person on the underlying (latent or hidden) indicators (for instance various respondents rate the importance of the 14 mobile phone attributes). It is used to weight the variables (mobile phone attributes). Distefano et al. (2009) in their article on Understanding and using Factor Scores: Considerations for the applied researcher pointed out that, in terms of loadings L , positive factor score correspond to lower than average ratings (or the respondent gave higher than average importance ratings), negative factor score correspond to higher than average ratings (or respondent gave lower than average importance ratings), while zero (0) factor score correspond to close average ratings (or the respondent ratings of the importance of a relevant attributes is close to the average for the sample). 39 University of Ghana http://ugspace.ug.edu.gh CHAPTER FOUR ANALYSIS AND RESULTS OF THE RESEARCH 4.0 INTRODUCTION This chapter involves analysis of the results of the study. It involves preliminary analysis (descriptive statistics- sample mean, sample standard deviation and bar graphs) and further analysis in which principal component (pc) method of factoring is used to obtain the set objectives. Statistical Package for Social Sciences (SPSS) is the software used for the analysis. A. PRELIMINARY ANALYSIS Data collected from the respondents are explored in this chapter to identify the nature of the variables. Through the exploration, some of the objectives and research questions would be answered. The mean and standard deviation of the attributes, age distribution, frequency distribution of mobile phone network, bar graphs of the opinion of respondents on the mobile phone attributes are used to explore the data. Bartlett‟s test of sphericity proposed by Bartlett (1954) as well as Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy proposed by Kaiser and Rice (1974) are used to ascertain the approprietness of the use of factor analysis in further analysis, while Cronbach‟s alpha is used to test the reliability (consistency) of the data (Nunnally,1978). 40 University of Ghana http://ugspace.ug.edu.gh 4.1 DISTRIBUTION OF AGE AND GENDER Table 4.1 shows the Age Distribution and Gender of the respondents obtained from Table A1 of Appendix II. Table 4.1: Distribution of Age and Gender of the Respondents. Age Male Female Total Total (%) Under 18 15 17 32 6.4 18-24 198 114 312 62.4 25-29 60 35 95 19.0 30-39 25 17 42 8.4 Over 39 10 9 19 3.8 Total 308 192 500 100.0 Table 4.1 shows the cross tabulation of age against gender; it can be seen from the table that 308 (constituting 61.6 %) males and 192 (constituting 38.4%) females were interviewed for the study, and these comprise students, lecturers and administrative staff (from Cape Coast Polytechnic). Moreover, majority (62.4%) of the respondents were aged between 18-24 in which 39.6% were males and 22.8% were females whilst very few (3.8 per cent) of the respondents were over 39 years. 41 University of Ghana http://ugspace.ug.edu.gh 4.2 THE MEAN AND STANDARD DEVIATION OF THE MOBILE PHONE ATTRIBUTES Table 4.2 shows the Mean and standard deviation of the mobile phone attributes obtained from Table A2 of Appendix II. Table 4.2: Mean and Standard Deviation of the variables Variable Mean Standard Deviation X1 3.99 1.16 X2 3.52 1.28 X3 3.29 1.38 X4 3.48 1.26 X5 3.70 1.21 X6 3.57 1.24 X7 3.68 1.89 X8 3.65 1.22 X9 3.60 1.27 X10 3.22 1.28 X11 3.91 1.07 X12 3.62 1.19 X13 3.57 1.21 X14 4.19 1.05 From Table 4.2, above, the highest mean (4.19) and the lowest standard deviation value is recorded by “X14 (long time usage)” of the network. This means the respondents consider the “long time usage” very important since this attribute had the highest rating. It is quite likely that a customer who has used a particular network for a long time would find it difficult to change over. The next attributes have high means which are less than what we have already seen, and also enjoy considerable amount of importance attached to them. These attributes are “X1 (wider coverage)” and “X11 (good advert)” with means 3.99 and 3.91 respectively. 42 University of Ghana http://ugspace.ug.edu.gh The last set of attributes that have lowest mean values, may be important to some people, they may not be as important as those in the first and second group, since these attributes received low ratings from the respondents. These attributes are “X3 (cheaper call cost)” and “X10 (games of chance)”, having means 3.29 and 3.22, respectively. It is interesting that almost all the variables have means around 3 with the exception of “X14 (long time usage)” which has a mean of 4.19. 4.3 USAGE OF MOBILE PHONE NETWORK This section shows the usage of the mobile phone network (MTN, Vodafone, Tigo, Airtel, Glo and Expresso) in the Cape Coast Polytechnic population. Tables 4.3 and 4.4 show the usage of the mobile phone network by the respondents, obtained from Table B1 to B6 of Appendix II. Table 4.3: Usage of Mobile Phone Network. Network Operator Usage MTN Tigo Vodafone Airtel Glo Expresso TOTAL Most Regular used 281 99 35 30 19 7 471 Regular used 22 65 51 28 21 6 193 Rarely used 28 51 46 25 24 1 175 Not used 1 7 12 13 15 12 60 Not Applicable 168 278 356 404 421 474 2101 TOTAL 500 500 500 500 500 500 3000 43 University of Ghana http://ugspace.ug.edu.gh Table 4.4: Usage of Mobile Phone Network by respondents in Percentage Network Operator (%) Usage MTN Tigo Vodafone Airtel Glo Expresso TOTAL Most Regular used 56.2 19.8 7.0 6.0 3.8 1.4 94.2 Regular used 4.4 13.0 10.2 5.6 4.2 1.2 38.4 Rarely used 5.6 10.2 9.2 5.0 4.8 0.2 35.0 Not used 0.2 1.4 2.4 2.6 3.0 2.4 12.0 Not Applicable 33.6 55.6 71.2 80.8 84.2 94.8 420.4 TOTAL 100.0 100.0 100.0 100.0 100.0 100.0 600.0 From Tables 4.3 and 4.4 above, convincing majority (303) of the respondents (constituting 60.6% from Table B1 of Appendix II) out of 500 respondents use MTN network regularly because, most of their friends and families are using the same network, faster internet speed, wider coverage and mobile phone banking services (MTN mobile money). Moreover, 5.6% of the respondents rarely use the network and only one (0.2%) respondent said he or she does not use the network at all. Overwhelmingly, 33.6% of the respondents said they are not connected to MTN network due to unreliable network. Moreover, the usage of Tigo (from Table B2 of Appendix II) network by Cape Coast Polytechnic population show that majority (55.6%) of the respondents are not Tigo subscribers, but 32.8% of the subscribers use the network regularly because it has faster internet speed, organize banking services called Tigo cash. But 1.2% of the Tigo subscribers have abandoned their chips. This is due to unreliable network or want to use same network with their friends and families. Furthermore, the Vodafone usage is very surprising. The majority (71.2%, Table B3 of Appendix II) of the respondents said they are not Vodafone subscribers. But 17.2% of the subscribers use this particular network regularly. Moreover, 2.4% of the respondents who are 44 University of Ghana http://ugspace.ug.edu.gh Vodafone subscribers are no more using the network. Their reason is to have a network that provides mobile phone banking services. With respect to the Airtel network, overwhelming majority (80.8%, from Table B4of Appendix II) of the respondents are not subscribers. This is due to lack of reception in their area of residence and also most of their friends and families are not using same network. Though, 2.6% subscribers are not using the network but only a handful (11.6%) use the Airtel network regularly. Besides, Glo network which has 8.0%, (Table B5 of Appendix II) of the market share in Cape Coast Polytechnic, use of the network regularly is extremely low. Surprisingly, the greater proportion 84.2% (Table B5 of Appendix II) of the respondents said they do not have interest in using the network at all because it has no wider coverage, do not operate mobile phone banking services and their friends are not using the same network. Some respondents also said that since they are already subscribers of other network, they find it difficult to change network. Three percent of the subscribers who decided to use the network have now abandoned their chips due to the reasons outlined above. Lastly, 474 (constituting 94.8%, from Table B6 of Appendix II) out of 500 respondents are not Expresso subscribers. A very small proportion (2.6 %) of the subscribers use the network regularly, which is extremely low compared to other networks, while 2.4% of the respondents who were once users are no more using the Expresso network, the reasons given by the respondents are lack of wider coverage, it is analog, do not provide mobile phone banking services and most of their friends and families are not using the network. 45 University of Ghana http://ugspace.ug.edu.gh University of Ghana http://ugspace.ug.edu.gh University of Ghana http://ugspace.ug.edu.gh 4.4.3 The Bar Graph of Lower Tariff (Cheaper Call Cost) Figure 3 shows the bar graph of the lower tariff (cheaper cost) obtained from Table C3 of Appendix II. FIGURE 3: Distribution of lower tariff (cheaper cost) From the bar graph above, most of the respondents (56.8%) agree that they would choose a network that has lower tariff. Surprisingly 35.4% of the respondents said lower tariff will not be their network attribute in selecting a mobile phone network. 48 University of Ghana http://ugspace.ug.edu.gh 4.4.4 The Bar Graph of Frequent Promotions and Motivations Figure 4 shows the bar graph of frequent promotions and Motivation packages of mobile network (obtained from Table C4 of Appendix II). FIGURE 4: Distribution of frequent promotions and Motivation packages of mobile phone network. From figure 4, it can be seen that the majority (62.0%) of the respondents agreed that the motivational and reward packages aspect influenced them in choosing a particular type of mobile phone network. But 26% of them said motivational aspect of the network will not lure them in selecting a network to be used. 49 University of Ghana http://ugspace.ug.edu.gh 4.4.5 The Bar Graph of Close Relations who use the same Network Figure 5 (obtained from Table C5 of Appendix II) shows the bar graph of Close relations (eg. friends, family) who use same network. FIGURE 5: Distribution of Close relations (eg. friends, family) who use the same network. The bar graph of figure 5 shows that the relationship with other people influences one‟s choice of the mobile phone network, since the greater proportion (71.0%) agreed. This means an individual would choose a particular network if a lover, family or friend uses same network. But 20.6% disagreed with this attribute. 50 University of Ghana http://ugspace.ug.edu.gh 4.4.6 The Bar Graph of Fast Internet Speed Figure 6 (obtained from Table C6 of Appendix II) shows the bar graph of the fast internet speed. FIGURE 6: Distribution of fast Internet Speed. It is clear from figure 6 that subscribers (65.8%) look for fast Internet speed before making a choice. There is no doubt that an individual would choose a network that has faster internet speed, because, internet technology plays a vital role in both information delivery and retrieval especially in e-learning, e-commerce and e-banking. 20.8% said they would not consider the internet speed in selecting a network to use. 51 University of Ghana http://ugspace.ug.edu.gh 4.4.7 The Bar Graph of Cheaper Starter Pack Figure 7 shows the bar graph of Cheaper Starter Pack obtained from Table C7 of Appendix II. FIGURE 7: Distribution of Cheaper starter pack. From Figure 7, the majority (68.2%) of the respondents said, there is no way that they will buy a chip that is very expensive. Thus, a new customer would definitely choose a network that has cheaper starter pack. But some (18.4%) respondents said they would not consider the price of the starter pack before becoming subscribers, while 13.4% of the respondents neither agreed nor disagreed to this network attribute. 52 University of Ghana http://ugspace.ug.edu.gh 4.4.8 The Bar Graph of Strong Network in my Area Figure 8 shows the bar graph of the strong area network obtained from Table C8 of Appendix II. FIGURE 8: Distribution of strong network in my area. It is clear from figure 8 that, most of the respondents (70.0%) agree that they would choose a network that has strong reception in their area. This is shown by the two highest bars in the graph. But 21.6% of the respondents disagree that strong area reception would influence them in choosing a network. 53 University of Ghana http://ugspace.ug.edu.gh 4.4.9 The Bar Graph of Mobile Phone Banking Services Figure 9 (obtained from Table C9 of Appendix II) shows the bar graph of Mobile phone banking services FIGURE 9: Distribution of Mobile Phone Banking service. The mobile phone banking services MTN mobile money, Tigo cash and Airtel money, operated by MTN, Tigo and Airtel respectively, have gained popularity in our society, since money can be sent and received through mobile phone. There is no doubt that overwhelming majority (69%) of the respondents said their selection of mobile phone network would be influenced by mobile phone banking; this is shown by figure 9. Nowadays, some small and 54 University of Ghana http://ugspace.ug.edu.gh University of Ghana http://ugspace.ug.edu.gh disagree that games of chance would influence them in any way. However, 19% of the respondents said they are neutral or undecided. 4.4.11 The Bar Graph of the network has Good Advert. Figure 11 (obtained from Table C11 of Appendix II) shows the bar graph of the Network has Good Advert. FIGURE 11: Distribution of the network has Good Advert. It is believed that “good advertisement” (the third most important attribute) plays an important role in marketing, to which the majority (77.0%) of the respondents agreed. Very good advertisement draws public attention to a particular network, especially funny advertisement would have greater influence on the public‟s choice. But only 12.8% of the 56 University of Ghana http://ugspace.ug.edu.gh University of Ghana http://ugspace.ug.edu.gh University of Ghana http://ugspace.ug.edu.gh since part of the company‟s profits, is re-invested in the society it is a good thing to be done. But 20.6% (Figure 13) respondents said they do not think that social or charity work of the operators would persuade them to choose a particular network. 4.4.14 The Bar Graph of long Time usage of a Network Figure 14 shows the bar graph of long time usage of a network obtained from Table C14 of Appendix II. FIGURE 14: Distribution of Long time usage of a network. Surprisingly, the vast majority 86.2% (Figure 14 obtained from Table C14 of Appendix II) of the respondents said since they have used a particular network for a “long time” (the most influential attribute) it would be very difficult to change over for another network. That is 59 University of Ghana http://ugspace.ug.edu.gh why the new operators (Airtel, Glo and Expresso) are finding it difficult to penetrate the network subscription market. But 10.4% of the respondents said they can easily switch to another network even if they have used the network for a long time. 4.5 CORRELATION MATRIX OF THE VARIABLES (MOBILE PHONE ATTRIBUTES) Table 4.5 shows the correlation matrix of the mobile phone attributes(variables) obtained from Table F1 of Appendix II. Table 4.5: Correlation Matrix. X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X1 1.000 X2 .444 1.000 X3 .111 .320 1.000 X4 .204 .343 .445 1.000 X5 .239 .288 .180 .368 1.000 X6 .310 .478 .343 .333 .248 1.000 X7 .304 .293 .347 .390 .289 .412 1.000 X8 .335 .432 .245 .355 .394 .397 .215 1.000 X9 .239 .253 .056* .222 .256 .222 .228 .344 1.000 X10 .170 .269 .330 .456 .349 .296 .286 .317 .298 1.000 X11 .198 .298 .301 .304 .290 .325 .420 .303 .241 .387 1.000 X12 .190 .222 .289 .379 .314 .340 .262 .333 .257 .402 .295 1.000 X13 .161 .283 .309 .390 .300 .366 .366 .331 .193 .308 .385 .484 1.000 X14 .305 .276 .102* .232 .264 .264 .298 .294 .281 .192 .284 .318 .350 1.000 *Value not significant at 0.05 60 University of Ghana http://ugspace.ug.edu.gh The correlation matrix of the data obtained to understand the attributes of mobile phone network is shown in Table 4.5 (obtained from Table F1 of Appendix II). Surprisingly, there are relatively low positive correlations among all the mobile phone attributes ranging from 0.111 to 0.484 . The lowest correlation 0.111 is between X1 (has wider coverage) against X3 (cheaper call cost). This convinces us that a network that has wider 60 coverage is not necessarily cheap or a network that is cheap does not mean it has wider coverage. Moreover, the highest correlation 0.484 is found between X12 (regular sponsorship of national events, eg. football) and X13 (encouraging charity work). It is not surprising that these two mobile phone attributes have relatively high correlation as compared to others, this means, social responsibility of the network operator influences one‟s choice of that particular network. This is because during these activities (events) adverts are made and it creates awareness so that when one wants to purchase a network, that particular network would be most appealing. This is confirmed by the bar graphs shown in Figure 12 and 13. The null hypothesis, H0 that the population correlation matrix is an identity matrix or the variables are uncorrelated r  0 - is rejected by Bartlett‟s test of sphericity. Since there are correlations (though, low positive correlations) among the mobile phone attributes, then there is a strong reason for factoring or factor analysis is an appropriate technique for the analysis. 61 University of Ghana http://ugspace.ug.edu.gh 4.6 TEST OF RELIABILITY (OR CONSISTENCY) Table 4.6 (obtained from Table E1 of Appendix II) shows the test of internal reliability (consistency) of the data. Table 4.6: Test of reliability (or consistency) of the collected data. Cronbach‟s alpha Number of items .857 14 The Table 4.6 gave Cronbach‟s alpha (coefficient) of 0.857 (Table E1 of Appendix II) which is greater than 0.700 (Nunnally, 1978). This means the data collected from the respondents is reliable (consistent) and can be used for the analysis. 4.7 KAISER-MEYER-OLKIN (KMO) AND BARTLETT’S TEST. Table 4.7 (obtained from Table E2 of Appendix II) shows the test values of Kaiser-Meyer- Olkin (KMO) and Bartlett‟s Test of sphericity. Table 4.7: Kaiser-Meyer-Olkin (KMO) and Bartlett‟s Test. Test Value Kaiser-Meyer-Olkin (KMO) 0.887 Bartlett‟s Test of Sphericity 1918.0 Degrees of freedom 91 Significance 0.000 62 University of Ghana http://ugspace.ug.edu.gh The Table 4.7 shows Kaiser-Meyer-Olkin (KMO) value of 0.887 (from Table 4.7) which is very close to 0.90, is large (that is > 0.5), we therefore say that the data is very adequate for factoring. This indicates that correlations between pairs of variables (mobile phone attributes) can be explained by other variables (mobile phone attributes) and that, factor analysis is appropriate and correlation matrix is appropriate for factoring (Kaiser & Rice, 1974). Moreover, the Bartlett‟s test of sphericity is also highly significant with p-value of 0.000 at relatively high chi-square value of 1918.00. These tests suggest that factor analysis is appropriate for the data B. FURTHER ANALYSIS This section concerns further analysis of the data collected from Cape Coast Polytechnic population. The factor analysis being the main statistical technique used in this research, is to find the underlying (latent) few factors that summarize the fourteen set of variables (mobile phone attributes). It was seen in the previous section that most of the variables are positively correlated and preliminary analysis generally revealed that factor analysis is suitable for the data as indicated by Bartlett‟s tests of sphericity and KMO value of sample adequacy. There is therefore the need to further investigate to identify the groupings (factors) among the attributes (variables) that influence subscribers‟ choice of mobile phone network. 63 University of Ghana http://ugspace.ug.edu.gh 4.8 NUMBER OF FACTORS TO EXTRACT In this section eigenvalues and scree plot is used to determine the number of factors that best summarizes the data. 4.8.1 Use of Eigen Value Analysis. In this, eigenvalues greater- than- one rule is used to determine the number of factors (component) to be extracted or retained (Zwich & Velicer, 1986; Cliff, 1988). Table 4.8 (obtained from Table G1 of Appendix) shows the eigenvalues (total variation explained) by the mobile phone attributes. Table 4.8: Total Variance Explained. Component Eigenvalues % of variance 1 4.948 35.35 2 1.251 8.94 3 1.064 7.60 4 0.927 6.62 5 0.820 5.86 6 0.737 5.27 7 0.680 4.86 8 0.623 4.45 9 0.608 4.35 10 0.539 3.85 11 0.507 3.62 12 0.489 3.49 13 0.423 3.02 14 0.381 2.72 The Table 4.8 (obtained from Table G1 of Appendix II) indicates the eigenvalues and percentage of variation explained by all the fourteen indicator variables (mobile phone 64 University of Ghana http://ugspace.ug.edu.gh attributes). It is clear from Table 4.8 that only three (3) components (mobile phone attributes) out of the original fourteen (14) components (mobile phone attributes) have eigenvalues 1  4.948 , 2 1.251and 3 1.064 significantly greater than one (Table 4.8). This suggests that three factors can be extracted. 4.8.2 Scree Plot Figure 15 shows a scree plot (a plot of eigenvalues against corresponding components number) obtained from Table G1 of Appendix II. Figure 15: Scree plot of eigenvalues and components. 65 University of Ghana http://ugspace.ug.edu.gh From the scree plot (figure 15), it can be seen that a distinct break occurs at the second component (factor), that is where an „elbow‟ of the diagram is shown. Moreover, from the second component (factor) on, we can see that the line is almost flat (uniform) meaning each successive factor is accounting for smaller and smaller amounts of the total variation. This pre-supposes that two factors can be considered for extraction. In all, the maximum number of factors that best explain the data set must not exceed three. 66 University of Ghana http://ugspace.ug.edu.gh 4.9 REPRODUCE CORRELATION MATRIX Table 4.9 shows the Reproduced Correlation Matrix and Residuals of the variables (mobile phone attributes) obtained from Table F2 of Appendix II. Table 4.9: Reproduce correlation matrix. X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14 X1 .640* X2 .595 .659* X3 .093 .350 .676* X4 .159 .337 .515 .539* X5 .269 .249 .164 .353 .408* X6 .450 .572 .454 .426 .285 .550* X7 .303 .445 .462 .449 .291 .470 .437* X8 .449 .462 .193 .345 .400 .417 .359 .483* X9 .326 .241 -.084 .191 .404 .192 .171 .418 .513* X10 .086 .212 .369 .485 .395 .308 .361 .335 .282 .499* X11 .205 .322 .379 .450 .352 .375 .384 .355 .252 .423 .395* X12 .095 .187 .291 .463 .433 .274 .331 .361 .355 .509 .420 .537* X13 .137 .263 .385 .491 .397 .346 .384 .359 .287 .492 .431 .498 .490* X14 .366 .321 .025 .244 .391 .274 .239 .432 .462 .294 .284 .346 .306 .438* X1 X2 -.152 X3 .018 -.029 X4 .045 .006 -.070 X5 -.006 .018 .016 .016 X6 -.140 -.094 -.111 -.093 -.037 X7 .000 -.152 -.114 -.059 -.002 -.058 X8 -.114 -.030 .052 .011 -.006 -.020 .057 X9 -.087 .012 .139 .032 -.148 .030 -.075 -.074 X10 .084 .056 -.039 -.029 -.046 -.012 .036 -.018 .016 X11 -.007 -.024 -.078 -.145 -.061 -.050 -.069 -.052 -.011 -.036 X12 .095 .035 -.002 -.084 -.119 .066 -.018 -.028 -.098 -.107 -.125 X13 .024 .020 -.078 -.101 -.097 .020 -.018 -.028 -.093 -.184 -.046 -.014 X14 -.062 -.046 .078 -.012 -.127 -.011 .059 -.138 -.181 -.101 .000 -.028 .044 67 University of Ghana http://ugspace.ug.edu.gh Table 4.9 contains two tables, the reproduced correlation matrix in the top part of the table, the diagonal values in the reproduced matrix (indicated by *), is the communalities; and the residuals (the differences between the observed correlations and the reproduced correlations), in the bottom part of the table for the fourteen (14) mobile phone attributes. From the table, the values of the reproduced matrix are close to the values in the original correlation matrix, or residual matrix is close to zero. Since the reproduce matrix is very similar to the observed correlation matrix, then the three factors that were extracted accounted for a great deal of the variance in the original (observed) correlation matrix. Thus, these few (three) factors do a good job of representing the original data. 4.10 UN-ROTATED FACTOR (COMPONENT) MATRIX Table 4.10 (obtained from Table G3 of Appendix II) shows the Un-rotated component (factor) matrix of the variables. Table 4.10: Un-rotated Component Matrix. Variable Factor 1 Factor 2 Factor 3 X1 .490 .563 -.288 X2 .628 .296 -.420 X3 .530 -.494 -.389 X4 .665 -.309 -.026 X5 .575 .073 .269 X6 .652 .055 -.348 X7 .618 -.107 -.209 X8 .642 .263 .045 X9 .473 .368 .393 X10 .614 -.265 .226 X11 .610 -.145 .043 X12 .618 -.194 .342 X13 .640 -.226 .171 X14 .525 .320 .246 68 University of Ghana http://ugspace.ug.edu.gh 4.11 VARIMAX ROTATED COMPONENT (FACTOR) MATRIX Table 4.11 (obtained from Table G4 of Appendix II) shows the Varimax Rotated Component (Factor) Matrix Table 4.11: Varimax Rotated Component Matrix. Variable Factor 1 Factor 2 Factor 3 X1 - .055 .735 .312 X2 .232 .763 .154 X3 .724 .291 -.259 X4 .688 .200 .159 X5 .353 .142 .513 X6 .421 .602 .100 X7 .511 .407 .103 X8 .266 .428 .478 X9 .072 .154 .696 X10 .621 .022 .336 X11 .533 .208 .261 X12 .573 -.021 .455 X13 .611 .093 .328 X14 .143 .258 .593 Considering the un-rotated factor matrix in table 4.10, and using a cut-off loading value of 0.5, you can see that there are high loadings for almost all the mobile phone attributes on factor one with exception of X8 (strong coverage in my area) and X9 (mobile phone banking), whereas only one attribute X8 (strong coverage in my area) loads high on factor two. It is interesting to note that since no attribute has high loading on factor three (3) this makes the interpretation of the factors very difficult. 69 University of Ghana http://ugspace.ug.edu.gh 4.11.1 Factor Rotation To determine the number of components that need to be retained we follow the rules of the “elbow” of the scree plot and eigenvalue-greater-than-one (Cliff, 1988). These rules suggest a maximum of three (3) factors. The varimax rotated factor structure of Table 4.11 is used for easy interpretation. Notice that the mobile phone attributes (indicators) can be written as a function of three common factors and fourteen (14) unique factors by: X1  0.055 f1 0.735 f2 0.312 f3 1 X2  0.232 f1 0.763 f2 0.154 f3 2 . . . . . . . . . X14  0.143 f1 0.258 f 0.593 f  2 3 14 The above equations can be written in the form X = LF + ε defined in equation 3.1. It can be seen from Table 4.8 that the variances accounted for by the three (3) factors f1, f2 and f3 are 1  4.948 , 2 1.251and3 1.064 respectively. The total amount of correlation among the indicators (mobile phone attributes) explained by the three factors is 7.263 4.9481.2511.064 which constitute 52.0% (from Table G1 of Appendix II) of the total variation. The unique variance ε accounts for the remaining 48.0%. The varimax rotated factor pattern of Table 4.11, proposed by Kaiser (1958) form the basis for easy interpretation of the factors. It is important to note that the factor rotation does not change the underlying solution. 70 University of Ghana http://ugspace.ug.edu.gh It can be seen from Table 5.4 that factor 1 has high coefficients (correlations or loadings) for attributes X3 (lower tariff), X4 (motivation), X7 (cheaper starter pack), X10 (organizes games of chance), X11 (has good advert), X12 (sponsorship of national events eg. football) and X13 (charity work), this factor can be labeled as social responsibility factor or customer benefit (care) factor. Thus, subscribers consider the reward schemes that the network operators can offer him or her and the society before choosing a particular network. It is convincing that a sports fan would definitely go in for a network operator that sponsors sports (football matches). Also, there is no doubt that an entertainment (music and dance) fan would choose a network that sponsors entertainment. It is interesting that factor 2 has high loadings or correlation on the attributes X1 (has wider coverage), X2 (has reliable network) and X6 (faster internet speed), this factor can be termed, reception benefit factor. There is no doubt that potential subscribers look for quality network before making a choice. Finally, factor 3 has high coefficients or loadings (correlation) for attributes X5 (close relations use same network), X9 (mobile phone banking) and X14 (used for long time) but the variable X8 (strong reception in my area), which is marginally low (slightly lower than 0.500) can be considered in factor three, therefore this factor may be labeled relationship benefit factor. It is convincing that subscribers choose network which has direct relation or benefit from his or her relationship with people he or she knows. 71 University of Ghana http://ugspace.ug.edu.gh 4.11.2 Final Factor Solution The solution obtained from the rotated factor matrix which gives us the final latent factors that best explain the variation in the choice of mobile phone network of the Cape Coast Polytechnic population. Notice that, the varimax transformation matrix is critical in determining the order of importance of the factors. From all the analysis carried out, it is clear that, the three (3) factor solution is appropriate and adequate in explaining the differences that exist in the choice of mobile phone network of the people of Cape Coast Polytechnics. The first factor is the social responsibility factor (or customer care factor) of the people, the next is the reception benefit factor, whilst the last factor is the relationship benefit factor. These three (3) factors identified, best summarize the people‟s choice of mobile phone network in Cape Coast Polytechnic. 4.12 ESTIMATING FACTOR SCORE Using Table G6 of Appendix II, component or factor score matrix of the mobile phone attributes, factor score can be estimated (using equation 3.16) as: f1  (.250) (.079) (.356) .198 .219 (.107) 1.066 f2  .466 .459 (.109) .238 .137 .022  .897 f3  .075 .105 (.403) (.216) .090 .344  .940 From the estimated factor score above it is interesting to note that factor one (social responsibility or customer care factor) had the highest score (1.066), factor three (relationship benefit factor) recorded the next highest score (0.940), while factor two (reception benefit factor) recorded the least score (0.896). This means that the respondents considered “social responsibility or customer care factor” very important or influential followed by the 72 University of Ghana http://ugspace.ug.edu.gh “relationship benefit factor”, while “reception benefit factor” is the least important or influential by Cape Coast Polytechnic population. In follow-up analysis (further analysis) these estimated factor scores can be used in regression analysis as well as analysis of variance (ANOVA). 73 University of Ghana http://ugspace.ug.edu.gh CHAPTER FIVE SUMMARY, CONCLUSION AND RECOMMENDATIONS 5.0 INTRODUCTION The results of the research are clearly stated and explained in this Chapter. The findings are summarized, while conclusions and recommendations are made. Suggestions for further studies are also embodied in this chapter. 5.1 SUMMARY OF FINDINGS Discoveries from this research were made based on the empirical analysis made in the previous chapter. Some of these findings are in line with other people‟s studies similar to mine. The research findings are enumerated as follows: 1. The respondents consider “long time usage” very important (from Table 4.2) since this attribute had the highest rating (with highest mean of 4.19), while “wider coverage” and “good advert” are considered the second most important attributes. 2. These attributes “lower tariff” and “games of chance” having means 3.29 and 3.22 (from Table 4.2) respectively, were considered as less important by the respondents. 3. The three most regularly used mobile phone network by Cape Coast Polytechnic population are MTN (60.6%), Tigo (32.8%) and Vodafone (17.2%) from Table 4.4. But Airtel and Globacom, recorded 11.6% and 8.0% respectively of the market share in Cape Coast Polytechnic, while the least patronized network is Expresso with 2.6% of the respondents (from Table 4.4). Notice that, due to multiple subscriptions, the sum of the percentages of the most regularly used networks would be more than 100. 4. The lowest correlation (0.111), from Table 4.5, is between X1 (wider coverage) against X3 (lower tariff). This tells us that a network that has wider coverage is not 74 University of Ghana http://ugspace.ug.edu.gh necessarily cheap or a network that is cheap does not mean it has wider coverage. Whilst the highest correlation (0.484), from Table 4.5, is found against X12 (regular sponsorship of national events. eg. football) and X13 (encouraging charity work). It is not surprising that these two mobile phone attributes have high correlation as compared to others, this means, social responsibility of the network operator influences one‟s choice of mobile phone network. 5. In all, three (3) factors were extracted, factor one (social responsibility factor or customer care factor), the next is factor two (reception benefit factor) and factor three is the relationship benefit factor of the people in Cape Coast Polytechnic. 6. The respondents considered “social responsibility or customer care factor” the most influential or important, followed by “relationship benefit factor”, while the “reception benefit factor” is the third important factor. 7. The total amount of correlation among the mobile phone attributes and the three factors extracted is 7.263 4.9481.2511.064, Table 4.8which constitute 52.0% of the total variation explained while the remaining 48.0% of the total variation is unexplained (that is the unique variance ε accounts for 48%). 5.2 DISCUSSION OF FINDINGS It is convincing to recognize that “long time usage” is the most important (86.2%, from Figure 14) attribute that the subscribers consider. It is true that a subscriber who has used a particular network for a very long time finds it very difficult to change over since most of his/her relations would miss him/her. The next most important attribute is “wider coverage” this is convincing that potential customers look for a network that has nationwide coverage (81.4%, from Figure 1), while “good advert” is the third most important attribute. It is 75 University of Ghana http://ugspace.ug.edu.gh believed that good and constant advertisement draws potential subscriber‟s attention to that network. The respondents consider “games of chance” as less important (49%, from Figure 10). The respondents foresee that lucky games (lotteries) organized by the network operators do not influence them in their choice of mobile phone network. The MTN which has largest market share as announced by the National Communication Authority (May 2015) in the country‟s Daily Graphic, still continues to enjoy that privilege in Cape Coast Polytechnic. Vodafone has the third market share after Tigo in the study area. These two mobile phone operators (MTN and Tigo) that have the largest market share in Cape Coast Polytechnic operate mobile money services-MTN mobile money and Tigo Cash. This might be the reason why MTN and Tigo have overtaken Vodafone (formerly Onetouch) in terms of market share. It will not be surprising if Airtel (which provides Airtel money services) overtake Vodafone in the near future. This is a wake-up call for the network providers who are not in mobile money banking services to do so. Expresso which has the least market share in the study area, operates analog services, that is subscribers do not use chip and one has no chance of changing the phone once they have bought analog phone. I believe that if they continue with this service their already existing subscribers might change for other networks. Though, correlations among the mobile phone attributes are low, the highest correlation is found between “regular sponsorship of national events, eg. football” and “encouraging charity work”. It is not surprising that these two mobile phone attributes have a high positive 76 University of Ghana http://ugspace.ug.edu.gh correlation as compared to others. This means, social responsibility of the network operator influences one‟s choice of mobile phone network. The lowest positive correlation recorded between “wider coverage” and “cheaper call cost” indicate that a network that has wider coverage is not necessarily cheap or a network that is cheap does not mean it has wider coverage. The three (3) factors that summarize the data showed that subscribers consider “social responsibility or customer care factor” (which talks about what the operators are doing for its customers and the nation at large) is considered the most important or influential by the respondents. It is convincing that what mobile phone operators are doing for the society is having a positive impact on the people of Cape Coast Polytechnic. The “relationship benefit factor” (which talks about the relationship with other people) is the next most important or influential factor. It is not surprising that very good friends, business partners and family members use the same network since the reward schemes and promotions (free calls to the same network) would be enjoyed by them. Thus, the relationship with other people influences one‟s choice of mobile phone network by the people in Cape Coast Polytechnic. The “reception benefit factor” is considered the third important or influential factor when making a choice. This is drawing our attention to the fact that “wider coverage”, “reliability” and “fast internet speed” of a network are not all that important when it comes to the choice of mobile phone network. 77 University of Ghana http://ugspace.ug.edu.gh 5.3 CONCLUSION AND RECOMMENDATIONS In conclusion, respondents who decided to subscribe to MTN network considered “wider coverage” as their most important attribute. Secondly, the respondents also consider “good advert” as their next most important attribute when it comes to subscription of a network. Moreover, majority of the respondents considered “mobile money services” as their third most important attribute that lure them in choosing MTN and Tigo network. The least important (or influential) attribute is “games of chance”. Furthermore, respondents consider “social responsibility or customer care” as the most important and “reception benefit” of the network operator as the least important factor when choosing a network. Based on the findings, the network operators (MTN, Vodafone, Tigo, Airtel, Globacom and Expresso) should consider the following recommendations: 1. The network operators should increase their coverage area to increase catchment area to attract more potential subscribers, since wider coverage is considered as one of the most important attributes. 2. It is worth noting that good advert plays an important role in marketing. Thus, the operators should modify their advertisement, where necessary, to draw attention or attract more potential subscribers. 3. It was revealed in the analysis that subscribers consider firstly, the social or corporate responsibility of the network operator before making a choice. Thus, the operator 78 University of Ghana http://ugspace.ug.edu.gh should intensify their social and corporate responsibilities (charity work, etc.) to attract more potential subscribers. 4. The respondents are of the view that the analog mobile phone, deter them from subscribing that particular network (eg. Expresso network). Thus, those operators especially Expresso should change from operating analog services. 5. The respondents in the study area said they have accepted mobile phone banking as their means of monetary transaction. It is therefore recommended that, those network operators who are providing this service should improve it, and those who are not providing the mobile money services should do so to increase their market share. 79 University of Ghana http://ugspace.ug.edu.gh REFERENCES Akakpo, J. (2008). Rural Access: Options and Challenges for Connectivity and Energy. Ghana International Institute for Communication and Development (IICD) and Ghana Information Network for Knowledge Sharing (GINK), 41- 43. nd Anderson, T.W. (1984). An Introduction to Multivariate Statistical Analysis (2 ed.). New York, USA: John Wiley. Ansah, N.T., Nortey, N. N., & Doku, A. K. (2013). Prediction of subscribers‟ Brand Switching behavior and ergodic market share of network service providers in Ghana. Research Journal in Engineering and Applied Sciences, 2(4), 298-303. Arthur, W. (1906). “Communication by wire and wireless”. The wonders of Telegraph and Telephone. The world’s work: A History of our time, XIII, 8408-8422. Baker, R., & Burton, H. (2000). The Gray Matter: The forgotten Story of the Telephone. Telepress, St. Joseph, MI 200.ISBN 0-615-11329-X. Bartlett, M.S. (1954). A Note on Multiplying Factors for various Chi-Squared Approximations. “Journal of the Royal Statistical Society” (B), 16, 296 -298. Bruce, A., & Robert, V. (1990). Alexander Graham Bell and the Conquest of Solitude. Ithaca: Cornell University Press. Cattell, R.B. (1966). “The Meaning and Strategic Use of Factor Analysis”. In R. B. Cattell (ed.). Handbook of Multivariate Experimental Psychology Rand McNally, USA: Chicago. Choo, S., & Mokhtarian, P.L. (2006). Telecommunication and Travel Demand and Supply; Aggregate Structural Equation Models for the USA. Journal of Transportation Research Part A: Policy and Practice, 41, 4-18. 80 University of Ghana http://ugspace.ug.edu.gh Cliff, N. (1988). “The Eigenvalue-Greater-than One Rule and the Reliability of Components”. Psychological Bulletin, 103, 3, 276-279. Coe, L., & Lewis, A. (1995). The Telephone and its Several Inventors: History of our time. McFarland, SA, North Carolina. ISBN 0-7864-0138-9. Psychological Bulletin, 103, 2, 276-279. rd Cochran, W.G. (1977). Sampling Techniques (3 ed.). New York. John Wiley & sons. nd Comrey, A.L., & Lee, H.B. (1992). A first course in factor analysis (2 ed.). Hillside, N.J.: Erlbaum. Daniel, P., & McVeigh, A. (2013). An early History of the Telephone: Robert Hookes Acoustic Experiments and Silent Inventions. USA, Colombia University. Dibner, B. (1959). The Atlantic Cable. Burndy Library Inc. Distefano, C., Zhu, M. & Mindrila, D. (2009). Understanding and using Factor Scores: Considerations for the Applied Researcher. Practical Assessment, Research and Evaluation, 14 (20). Duncan, O. D. (1975). Introduction to Structural Equation Models. New York, USA: New York Academic Press. Evenson, A., & Edward, M. (2000). The Telephone Patent Conspiracy of 1876: The Elisha Gray-Alexander Graham Bell Controversy. McFarland, USA: North Carolina. ISBN 07864-0883-9. Grice, J.W. (2001). A comparison of factor scores under conditions of factor obliquity. Psychological Methods, 6, 67-83. Harman, H. H. (1976). Modern Factor Analysis. Chicago, USA: University of Chicago Press. Huurdeman, H., & Anton, A. (2003). The worldwide History of Telecommunications. 81 University of Ghana http://ugspace.ug.edu.gh IEEE Press and John Wiley & Sons, ISBN 0- 471-20505-2. In R. B. Cattell (ed.). Handbook of Multivariate Experimental Psychology Rand McNally, USA: Chicago. rd Johnson, R.A., & Wichern, D.W. (1992). “Applied Multivariate Statistical Analysis” (3 ed.). Prentice Hall International Inc. Jones, R., & Sommerings, V.S.T. (1965). “Space Multiplexed Electrochemical Telegraph. Harvard University. Kaiser, H. F. (1958). “The Varimax Criterion for Analytic Rotation in Factor Analysis”. Psychometrika, 23, 187-200. Kaiser, H. F., & Rice, J. (1974). “Little Jiffy Mark IV”, Educational and Psychological Measurement 34 (Spring), 111-117. Kolger, F., & Jon, G. (1986). “Mechanical or String Telephone”. ATCA Newsletter. Krejcie, R. V., & Morgan, D. W. (1970). Determining Sample size for research activities. Educational and Psychological Measurement, 30, 607-610. nd Lawley, D.N., & Maxwell, A.E. (1971). Factor Analysis as a Statistical Method (2 ed.). New York, American Elsevier Publishing Co. Lawrence Erlbaum, Hillsdale, N.J. Lindeman, W. (1937). “On the Rank of Reduced Correlation Matrix in Multiple Factor Analysis”. Psychometrica, 2, 85-99. Maxwell, A. E., & Lawley D. N. (1971). Multivariate Analysis in Behavioral Research. London; Chapman and Hall. Miller, C.E. (1980). Telecommunication/Transportation Substitution; Some Empirical Findings. Journal of Socio-Economic Planning Sciences, 14, 163-166. nd Morrison, D. F. (1976). Multivariate Statistical Methods (2 ed.) New York, McGraw- Hill. Multiple Factor Analysis”. Psychometrica 2, 85- 99. 82 University of Ghana http://ugspace.ug.edu.gh th Naresh, R. (2004). Marketing Research; An Applied Orientation (4 ed.). Georgia, USA: Georgia Institute of Technology. New York, USA. nd Nunnally, J. (1978). Psychometric Theory (2 ed.) New York. McGraw-Hill. Price Waterhouse Coopers. (2009). Survey of communication Chief Executive Officers th (C.E.O.’s): 15 Annual Report. Ren, F., & Kwan, M. (2008). The Impact of Internet on Human Activity-Travel Patterns: Analysis of Gender Differences Using Multi-grouped, Structural Equation Models. Journal of Transport Geography, 17, 440-450. Richard, A., & John, P. (2010). Network Nation: Inventing American Telecommunications. Harvard University Press. Rommel, R. J. (1970). Applied Factor Analysis. North Western University Press, Evanston III. Sharma, S. (1996). Applied Multivariable Techniques. John Wiley & Sons, USA. Spearman, C. (1909). A General Intelligence Objectivity”. Objectivity Determined and Measured”. America Journal of Psychology, 1909, 15, 201- 293. Stevens, S.S. (1946). “On the Theory of Scale Measurement”. Science Journal, 103, 677- 680. The Elisha Gray-Alexander Graham Bell Controversy. McFarland, USA: North Carolina. ISBN 07864-0883-9. Wheen, R., & Andrew, M. (2011). How Modern Telecommunications Evolved from the Telegraph to the Internet. (springer, 2011), ISBN 978-1-4419-6759-6. World Bank. (2013). Ghana Telecommunications Sector Investment Project. World Bank Project Report. Washington, DC. Zwick, W. K., & Velicer, W. F. (1986). Comparison of Five Rules for Determining the Number of Components to Retain. Psychological Bulletin, 99 (3), 432-442. 83 University of Ghana http://ugspace.ug.edu.gh APPENDIX I METHODIST UNIVERSITY COLLEGE GHANA DEPARTMENT OF MATHEMATICAL SCIENCES QUESTIONNAIRE QUESTIONNAIRE ON MOBILE PHONE NETWORK PREFERENCE STUDY The objective of this questionnaire is to determine the underlying factors that influence the customer‟s choice of a mobile phone network. The information that will be provided will be kept confidential. PART I Please tick [ √ ] where appropriate 1. Gender Code Male [ ] 1 Female [ ] 2 2. Age Under 18 years [ ] 1 18-24 years [ ] 2 25-29 years [ ] 3 30-39years [ ] 4 Over 39years [ ] 5 84 University of Ghana http://ugspace.ug.edu.gh 3. How many networks are you using now ……………………………………………… 4. Please mention them…………………………………………………………………… 5. Please, using the table below indicate by ticking [ √ ] the mobile phone network you use regularly. Network Most Regularly Rarely Not in Not Provider Regularly Used Used use Applicable Used (1) (2) (3) (5) (4) MTN VODAFONE TIGO AIRTEL GLO EXPRESSO 85 University of Ghana http://ugspace.ug.edu.gh PART II In relation to your choice of mobile phone network in Part I indicate your option on each of the following attributes. Please tick [ √ ] the appropriate cell in each column. Attributes of mobile Strongly Disagree Undecided Agree Strongly phone network Disagree Agree (1) (2) (3) (4) (5) X1. Network has wider coverage. X2. Network is reliable. X3. Lower tariff. X4. Frequent Promotions/Rewarding/ Motivation. X5. Close relations (eg. Friends, Family etc) use same network. X6. Faster internet speed X7. Cheaper starter pack X8. Strong network in my area X9. Mobile Phone Banking services X10. Frequently organize games of chance X11. Has good adverts X12. Regular sponsorship of national events (eg. football) X13. Active in charity/social work X14.Used for a long time Thank you 86 University of Ghana http://ugspace.ug.edu.gh APPENDIX II A1 Age * Gen crosstab Count Gen Total Male Female Under 18 15 17 32 18 -24 198 114 312 Age 25-29 60 35 95 30-39 25 17 42 over 39 10 9 19 Total 308 192 500 A2 Descriptive Statistics N Minimum Maximum Mean Std. Skewness Kurtosis Deviation Statistic Statistic Statistic Statistic Statistic Statistic Std. Error Statistic Std. Error Coverage 500 1.00 5.00 3.9920 1.16389 -1.355 .109 1.047 .218 Reliable 500 1.00 5.00 3.5160 1.28103 -.659 .109 -.752 .218 Lower tariff 500 1.00 5.00 3.2900 1.37757 -.356 .109 -1.240 .218 Motivation 500 1.00 5.00 3.4800 1.25569 -.598 .109 -.762 .218 Relationship 500 1.00 5.00 3.7000 1.20869 -.871 .109 -.248 .218 Internet Faster 500 1.00 5.00 3.5680 1.24438 -.710 .109 -.585 .218 Cheaper Pack 500 1.00 5.00 3.6760 1.18905 -.854 .109 -.149 .218 Area Reception 500 1.00 5.00 3.6540 1.21706 -.840 .109 -.324 .218 Mobile Banking 500 1.00 5.00 3.5960 1.26965 -.842 .109 -.407 .218 Games of Chance 500 1.00 5.00 3.2240 1.28262 -.270 .109 -1.058 .218 GoodAdvert 500 1.00 5.00 3.9060 1.06745 -1.132 .109 .784 .218 Sponsorship 500 1.00 5.00 3.6200 1.19266 -.770 .109 -.338 .218 Charity 500 1.00 5.00 3.5720 1.21316 -.703 .109 -.443 .218 Long Time 500 1.00 5.00 4.1940 1.05003 -1.593 .109 2.053 .218 Valid N (listwise) 500 87 University of Ghana http://ugspace.ug.edu.gh Frequency Table B1 MTN Cumulative Frequency Percent Valid Percent Percent Valid Most Regularly Used 281 56.2 56.2 56.2 Regularly Used 22 4.4 4.4 60.6 Rarely Used 28 5.6 5.6 66.2 Not Used 1 .2 .2 66.4 N/A 168 33.6 33.6 100.0 Total 500 100.0 100.0 B2 Tigo Cumulative Frequency Percent Valid Percent Percent Valid Most Regularly Used 99 19.8 19.8 19.8 Regularly Used 65 13.0 13.0 32.8 Rarely Used 51 10.2 10.2 43.0 Not Used 7 1.4 1.4 44.2 N/A 278 55.6 55.6 100.0 Total 500 100.0 100.0 88 University of Ghana http://ugspace.ug.edu.gh B3 Vodafone Cumulative Frequency Percent Valid Percent Percent Valid Most Regularly Used 35 7.0 7.0 7.0 Regularly Used 51 10.2 10.2 17.2 Rarely Used 46 9.2 9.2 26.4 Not Used 12 2.4 2.4 28.8 N/A 356 71.2 71.2 100.0 Total 500 100.0 100.0 B4 Airtel Cumulative Frequency Percent Valid Percent Percent Valid Most Regularly Used 30 6.0 6.0 6.0 Regularly Used 28 5.6 5.6 11.6 Rarely Used 25 5.0 5.0 16.6 Not Used 13 2.6 2.6 19.2 N/A 404 80.8 80.8 100.0 Total 500 100.0 100.0 B5 Glo Cumulative Frequency Percent Valid Percent Percent Valid Most Regularly Used 19 3.8 3.8 3.8 Regularly Used 21 4.2 4.2 8.0 Rarely Used 24 4.8 4.8 12.8 Not Used 15 3.0 3.0 15.8 N/A 421 84.2 84.2 100.0 Total 500 100.0 100.0 89 University of Ghana http://ugspace.ug.edu.gh B6 Expresso Cumulative Frequency Percent Valid Percent Percent Valid Most Regularly Used 7 1.4 1.4 1.4 Regularly Used 6 1.2 1.2 2.6 Rarely Used 1 .2 .2 2.8 Not Used 12 2.4 2.4 5.2 N/A 474 94.8 94.8 100.0 Total 500 100.0 100.0 Frequency Table C1 Coverage Cumulative Frequency Percent Valid Percent Percent Valid Strongly Disagree 37 7.4 7.4 7.4 Disagree 31 6.2 6.2 13.6 Undecided 25 5.0 5.0 18.6 Agree 213 42.6 42.6 61.2 Strongly Agree 194 38.8 38.8 100.0 Total 500 100.0 100.0 90 University of Ghana http://ugspace.ug.edu.gh C2 Reliable Cumulative Frequency Percent Valid Percent Percent Valid Strongly Disagree 50 10.0 10.0 10.0 Disagree 82 16.4 16.4 26.4 Undecided 43 8.6 8.6 35.0 Agree 210 42.0 42.0 77.0 Strongly Agree 115 23.0 23.0 100.0 Total 500 100.0 100.0 C3 Lower tariff Cumulative Frequency Percent Valid Percent Percent Valid Strongly Disagree 69 13.8 13.8 13.8 Disagree 108 21.6 21.6 35.4 Undecided 39 7.8 7.8 43.2 Agree 177 35.4 35.4 78.6 Strongly Agree 107 21.4 21.4 100.0 Total 500 100.0 100.0 91 University of Ghana http://ugspace.ug.edu.gh C4 Motivation Cumulative Frequency Percent Valid Percent Percent Valid Strongly Disagree 47 9.4 9.4 9.4 Disagree 83 16.6 16.6 26.0 Undecided 60 12.0 12.0 38.0 Agree 203 40.6 40.6 78.6 Strongly agree 107 21.4 21.4 100.0 Total 500 100.0 100.0 C5 Relationship Cumulative Frequency Percent Valid Percent Percent Valid Strongly Disagree 37 7.4 7.4 7.4 Disagree 66 13.2 13.2 20.6 Undecided 42 8.4 8.4 29.0 Agree 220 44.0 44.0 73.0 Strongly Agree 135 27.0 27.0 100.0 Total 500 100.0 100.0 92 University of Ghana http://ugspace.ug.edu.gh C6 Internet Faster Cumulative Frequency Percent Valid Percent Percent Valid Strongly Disagree 44 8.8 8.8 8.8 Disagree 75 15.0 15.0 23.8 Undecided 52 10.4 10.4 34.2 Agree 211 42.2 42.2 76.4 Strongly Agree 118 23.6 23.6 100.0 Total 500 100.0 100.0 C7 Cheaper Pack Cumulative Frequency Percent Valid Percent Percent Valid Strongly Disagree 39 7.8 7.8 7.8 Disagree 53 10.6 10.6 18.4 Undecided 67 13.4 13.4 31.8 Agree 213 42.6 42.6 74.4 Strongly Agree 128 25.6 25.6 100.0 Total 500 100.0 100.0 93 University of Ghana http://ugspace.ug.edu.gh C8 Area Reception Cumulative Frequency Percent Valid Percent Percent Valid Strongly Disagree 40 8.0 8.0 8.0 Disagree 68 13.6 13.6 21.6 Undecided 42 8.4 8.4 30.0 Agree 225 45.0 45.0 75.0 Strongly Agree 125 25.0 25.0 100.0 Total 500 100.0 100.0 C9 Mobile Banking Cumulative Frequency Percent Valid Percent Percent Valid Strongly Disagree 55 11.0 11.0 11.0 Disagree 57 11.4 11.4 22.4 Undecided 43 8.6 8.6 31.0 Agree 225 45.0 45.0 76.0 Strongly Agree 120 24.0 24.0 100.0 Total 500 100.0 100.0 94 University of Ghana http://ugspace.ug.edu.gh C10 Game of Chance Cumulative Frequency Percent Valid Percent Percent Valid Strongly Disagree 60 12.0 12.0 12.0 Disagree 100 20.0 20.0 32.0 Undecided 95 19.0 19.0 51.0 Agree 158 31.6 31.6 82.6 Strongly Agree 87 17.4 17.4 100.0 Total 500 100.0 100.0 C11 Good Advert Cumulative Frequency Percent Valid Percent Percent Valid Strongly Disagree 23 4.6 4.6 4.6 Disagree 41 8.2 8.2 12.8 Undecided 49 9.8 9.8 22.6 Agree 234 46.8 46.8 69.4 Strongly Agree 153 30.6 30.6 100.0 Total 500 100.0 100.0 95 University of Ghana http://ugspace.ug.edu.gh C12 Sponsorship Cumulative Frequency Percent Valid Percent Percent Valid Strongly Disagree 38 7.6 7.6 7.6 Disagree 64 12.8 12.8 20.4 Undecided 66 13.2 13.2 33.6 Agree 214 42.8 42.8 76.4 Strongly Agree 118 23.6 23.6 100.0 Total 500 100.0 100.0 C13 Charity Cumulative Frequency Percent Valid Percent Percent Valid Strongly Disagree 43 8.6 8.6 8.6 Disagree 60 12.0 12.0 20.6 Undecided 82 16.4 16.4 37.0 Agree 198 39.6 39.6 76.6 Strongly Agree 117 23.4 23.4 100.0 Total 500 100.0 100.0 96 University of Ghana http://ugspace.ug.edu.gh C14 Long Time Cumulative Frequency Percent Valid Percent Percent Valid Strongly Disagree 21 4.2 4.2 4.2 Disagree 31 6.2 6.2 10.4 Undecided 17 3.4 3.4 13.8 Agree 192 38.4 38.4 52.2 Strongly Agree 239 47.8 47.8 100.0 Total 500 100.0 100.0 E1 Reliability Statistics Cronbach's N of Alpha Items .857 14 E2 KMO and Bartlett's Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy. .887 Bartlett's Test of Sphericity Approx. Chi-Square 1.918E3 d f 91 Sig. .000 97 University of Ghana http://ugspace.ug.edu.gh F1 Correlation Matrix Lower Internet Cheaper AreaRece MobileBa GameOf Good Sponsor Coverage Reliable Tariff Motivation Relationship Faster Pack ption nking Chance Advert ship Charity LongTime Correlation Coverage 1.000 .444 .111 .204 .239 .310 .302 .335 .239 .170 .198 .190 .161 .305 Reliable .444 1.000 .320 .343 .288 .478 .293 .432 .253 .269 .298 .222 .283 .276 Lower Tariff .111 .320 1.000 .445 .180 .343 .347 .245 .056 .330 .301 .289 .307 .102 Motivation .204 .343 .445 1.000 .368 .333 .390 .355 .222 .456 .304 .379 .390 .232 Relationship .239 .288 .180 .368 1.000 .248 .289 .394 .256 .349 .290 .314 .300 .264 InternetFaster .310 .478 .343 .333 .248 1.000 .412 .397 .222 .296 .325 .340 .366 .264 CheaperPack .302 .293 .347 .390 .289 .412 1.000 .215 .228 .286 .420 .262 .366 .298 Area .335 .432 .245 .355 .394 .397 .215 1.000 .344 .317 .303 .333 .331 .294 Reception Mobile .239 .253 .056* .222 .256 .222 .228 .344 1.000 .298 .241 .257 .193 .281 Banking GameOfChance .170 .269 .330 .456 .349 .296 .286 .317 .298 1.000 .387 .402 .308 .192 GoodAdvert .198 .298 .301 .304 .290 .325 .420 .303 .241 .387 1.000 .295 .385 .284 Sponsorship .190 .222 .289 .379 .314 .340 .262 .333 .257 .402 .295 1.000 .484 .318 Charity .161 .283 .307 .390 .300 .366 .366 .331 .193 .308 .385 .484 1.000 .350 LongTime .305 .276 .102* .232 .264 .264 .298 .294 .281 .192 .284 .318 .350 1.000 *value not significant at 0.05 98 University of Ghana http://ugspace.ug.edu.gh F2 Reproduced Correlations Relati Mobile Game Spons Lon Covera Reliab Lower Motiv on Internet Cheaper AreaRec Bankin OfCha Good or Charit gTi ge le Tariff ation ship Faster Pack eption g nce Advert ship y me a Reproduced Correlation Coverage .640 .595 .093 .159 .245 .450 .303 .449 .326 .086 .205 .095 .137 .366 a Reliable .595 .659 .350 .337 .269 .572 .445 .462 .241 .212 .322 .187 .263 .321 Lower a .093 .350 .676 .515 .164 .454 .462 .193 -.084 .369 .379 .291 .385 .025 Tariff a Motivation .159 .337 .515 .539 .353 .426 .449 .345 .191 .485 .450 .463 .491 .244 Relationshi a .245 .269 .164 .353 .408 .285 .291 .400 .404 .395 .352 .433 .397 .391 p InternetFas a .450 .572 .454 .426 .285 .550 .470 .417 .192 .308 .375 .274 .346 .274 ter CheaperPa a .303 .445 .462 .449 .291 .470 .437 .359 .171 .361 .384 .331 .384 .239 ck AreaRecep a .449 .462 .193 .345 .400 .417 .359 .483 .418 .335 .355 .361 .359 .432 tion MobileBan a .326 .241 -.084 .191 .404 .192 .171 .418 .513 .282 .252 .355 .287 .462 king GameOfC a .086 .212 .369 .485 .395 .308 .361 .335 .282 .499 .423 .509 .492 .294 hance GoodAdve a .205 .322 .379 .450 .352 .375 .384 .355 .252 .423 .395 .420 .431 .284 rt 99 University of Ghana http://ugspace.ug.edu.gh Sponsorshi a .095 .187 .291 .463 .433 .274 .331 .361 .355 .509 .420 .537 .498 .346 p a Charity .137 .263 .385 .491 .397 .346 .384 .359 .287 .492 .431 .498 .490 .306 a LongTime .366 .321 .025 .244 .391 .274 .239 .432 .462 .294 .284 .346 .306 .438 b Residual Coverage - -.152 .018 .045 -.006 -.140 .000 -.114 -.087 .084 -.007 .095 .024 .062 Reliable - -.152 -.029 .006 .018 -.094 -.152 -.030 .012 .056 -.024 .035 .020 .046 Lower .018 -.029 -.070 .016 -.111 -.114 .052 .139 -.039 -.078 -.002 -.078 .078 Tariff Motivation - .045 .006 -.070 .016 -.093 -.059 .011 .032 -.029 -.145 -.084 -.101 .012 Relationshi - -.006 .018 .016 .016 -.037 -.002 -.006 -.148 -.046 -.061 -.119 -.097 p .127 InternetFas - -.140 -.094 -.111 -.093 -.037 -.058 -.020 .030 -.012 -.050 .066 .020 ter .011 CheaperPa .000 -.152 -.114 -.059 -.002 -.058 -.145 .057 -.075 .036 -.069 -.018 .059 ck AreaRecep - -.114 -.030 .052 .011 -.006 -.020 -.145 -.074 -.018 -.052 -.028 -.028 tion .138 MobileBan - -.087 .012 .139 .032 -.148 .030 .057 -.074 .016 -.011 -.098 -.093 king .181 GameOfC - .084 .056 -.039 -.029 -.046 -.012 -.075 -.018 .016 -.036 -.107 -.184 hance .101 100 University of Ghana http://ugspace.ug.edu.gh GoodAdve -.007 -.024 -.078 -.145 -.061 -.050 .036 -.052 -.011 -.036 -.125 -.046 .000 rt Sponsorshi - .095 .035 -.002 -.084 -.119 .066 -.069 -.028 -.098 -.107 -.125 -.014 p .028 Charity .024 .020 -.078 -.101 -.097 .020 -.018 -.028 -.093 -.184 -.046 -.014 .044 LongTime -.062 -.046 .078 -.012 -.127 -.011 .059 -.138 -.181 -.101 .000 -.028 .044 Extraction Method: Principal Component Analysis. a. Reproduced communalities b. Residuals are computed between observed and reproduced correlations. There are 45 (49.0%) non-redundant residuals with absolute values greater than 0.05. 101 University of Ghana http://ugspace.ug.edu.gh G1 Total Variance Explained Compo Initial Eigenvalues Rotation Sums of Squared Loadings nent Total % of Variance Cumulative % Total % of Variance Cumulative % 1 4.948 35.346 35.346 3.085 22.033 22.033 2 1.251 8.938 44.285 2.121 15.152 37.185 3 1.064 7.602 51.887 2.058 14.702 51.887 4 .927 6.624 58.511 5 .820 5.859 64.370 6 .737 5.268 69.638 7 .680 4.861 74.498 8 .623 4.447 78.945 9 .608 4.346 83.292 10 .539 3.853 87.144 11 .507 3.621 90.765 12 .489 3.491 94.256 13 .423 3.023 97.280 14 .381 2.720 100.000 Extraction Method: Principal Component Analysis. 102 University of Ghana http://ugspace.ug.edu.gh G2 a Un-rotated Component Matrix Component 1 2 3 Coverage .490 .563 -.288 Reliable .628 .296 -.420 Lower Tariff .530 -.494 -.389 Motivation .665 -.309 -.026 Relationship .575 .073 .269 InternetFaster .652 .055 -.348 CheaperPack .618 -.107 -.209 AreaReception .642 .263 .045 MobileBanking .473 .368 .393 GameOfChance .614 -.265 .226 GoodAdvert .610 -.145 .043 Sponsorship .618 -.194 .342 Charity .640 -.226 .171 LongTime .525 .320 .246 Extraction Method: Principal Component Analysis. a. 3 components extracted. 103 University of Ghana http://ugspace.ug.edu.gh G3 a Rotated Component Matrix Component 1 2 3 Coverage -.055 .735 .312 Reliable .232 .763 .154 Lower tariff .724 .291 -.259 Motivation .688 .200 .159 Relationship .353 .142 .513 Internet Faster .421 .602 .100 Cheaper Pack .511 .407 .103 Area Reception .266 .428 .478 Mobile Banking .072 .154 .696 Game Of Chance .621 .022 .336 Good Advert .533 .208 .261 Sponsorship .573 -.021 .455 Charity .611 .093 .328 Long Time .143 .258 .593 Extraction Method: Principal Component Analysis. Rotation Method: Varimax with Kaiser Normalization. a. Rotation converged in 16 iterations. G4 Component Transformation Matrix Component 1 2 3 1 .704 .510 .494 2 -.710 .506 .490 3 .000 -.696 .718 Extraction Method: Principal Component Analysis. Rotation Method: Varimax with Kaiser Normalization. 104 University of Ghana http://ugspace.ug.edu.gh G5 a Coefficient Matrix Component 1 2 3 Coverage .087 .311 -.301 Reliable .142 .188 -.486 Lower tariff .145 -.599 -.300 Motivation .146 -.226 .098 Relationship .113 .117 .224 InternetFaster .138 .022 -.322 CheaperPack .117 -.077 -.108 AreaReception .129 .216 -.023 MobileBanking .103 .404 .313 GameOfChance .140 -.137 .366 GoodAdvert .091 -.038 .056 Sponsorship .118 -.066 .298 Charity .126 -.109 .161 LongTime .073 .163 .054 Extraction Method: Principal Component Analysis. a. Coefficients are standardized. 105 University of Ghana http://ugspace.ug.edu.gh G6 Component Score Coefficient Matrix Component 1 2 3 Coverage -.250 .466 .075 Reliable -.079 .459 -.105 Lower Tariff .356 .109 -.403 Motivation .270 -.040 -.072 Relationship .040 -.088 .268 Internet Faster .062 .317 -.148 Cheaper Pack .149 .158 -.121 Area Reception -.058 .143 .197 Mobile Banking -.141 -.059 .456 Games of Chance .238 -.192 .110 Good Advert .169 -.024 .033 Sponsorship .198 -.238 .216 Charity .219 -.137 .090 Long Time -.107 .022 .344 Extraction Method: Principal Component Analysis. Rotation Method:Varimax with Kaiser Normalization. Factor Scores Method: Regression. 106 University of Ghana http://ugspace.ug.edu.gh 1