113 Upon the experts’ agreement to take part in the content validity, the survey form of the content validity was distributed to them (Appendix E). The survey form also provided instructions to the experts explaining the expert’s role and scoring measurement. The experts were asked to evaluate the relevancy of each item by providing their rating for each item based on five-point scale: “1=strongly disagree, 2=disagree, 3=agree (but not important), 4=agree, 5=strongly agree” (Allahyari et al., 2011; Allen & Seaman, 2007). The experts were also asked to give any comment or feedback on each construct measurement. Upon the completion completed the content validity survey by the experts, they were required to sign the confirmation letter which indicates their participation in the content validity of the research. Appendix F presents a sample of the expert involvement confirmation in content validity process. Following the analysis of experts’ feedback, a qualitative and quantitative analysis was performed. Qualitative analysis involves the subjective feedback from the experts on the measurement scale. Due to a large pool of items in the survey, the quantitative analysis also has been conducted to quantify the analysis (Straub, Boudreau, & Gefen, 2004). Quantitative analysis includes Content Validity Ratio (CVR) and Content Validity Index (CVI) calculation to measure the validity of the survey items (Wilson, Pan, & Schumsky, 2012). CVR is an item’s statistic indicating the usefulness of item measurement to be accepted of rejected. Using Lawshe (1975) content validity calculation, CVR was calculated for each measurement item in accordance with CVR calculation which is defined as follow: CVR = (Ne – N) / N The value Ne is the number of experts indicating “relevant” (score of 4 and 5), and the value N is the total number of experts. Based on the total number of experts which is eleven, minimum CVR of 0.59 is required to accept the measurement item to be retained in the survey Lawshe (1975). Appendix G shows the CVR result for all the measurement items. From the result, all measurement items were retained since the CVR values score were minimum 0.82, hence more than 0.59.
114 The CVR values of the content validity indicate that the experts agreed with all 52 proposed measurement items for the eleven constructs of the conceptual model. Based on the expert suggestions, some measurement items were refined in terms of language editing and formatting to suit the survey instrument in the research context. The experts agreed that the relative advantage consists of five measurement items. Four measurement items were adopted from Premkumar and Roberts (1999) which assess the potential advantages of MDM to improve the communication between the organization and other organizations, reduce data management cost, improve service delivery and provide timely decision-making. In addition, one item of relative advantage also aims to measure whether MDM could reduce the data quality issues in the organization or not (Vilminko-Heikkinen & Pekkola, 2013). The experts decided that the complexity would be measured using four measurement items. Two items of complexity to assess the organization’s difficulty in identifying potential master data for data sharing, and the complexity of data cleaning process before the data can be shared to MDM repository (Loshin, 2009). In addition to two more items of complexity to assess the difficulty in integrating MDM in the organization’s current work practices and skills required for MDM operation (Premkumar & Roberts, 1999). The experts agreed that the quality of master data consists of six measurement items. Therefore, the quality of master data was measured using these six items adopted from the DAMA UK Working Group (2013). The measurement of quality master data involves the assessment of the degree of master data completeness, uniqueness, timeliness, validity, accuracy, and consistency at the organization. The experts approved that the data security consists of five measurement items. The data security items evaluate the secure communication medium of the MDM channel, user authorization, and access rights to the data (Soliman & Janz, 2004). Additionally, two items of physical security of the data center of MDM, and a digital signature of the data exchange transaction were also approved as data security measurement items (Hristidis, Chen, Li, Luis, & Deng, 2010; Smallwood, 2014).
115 The experts concluded that the data governance consists of six measurement items. Five data governance measurement items were derived from Hung et al. (2014) to assess the MDM outcome analysis by the organizations, ongoing responsibilities taken, accountability, systematic procedures, business case definition and impact evaluation of the MDM. One more item of data governance which include the identification of stakeholder’s organization, data owner, and data stewardship (Smallwood 2014) was agreed to be retained in the survey. The experts agreed that the top management support consists of four measurement items. All four items were adapted from Premkumar and Roberts (1999). The measurement items include evaluating the interest, awareness, resource allocation, and vision of top management of the organization. The experts approved that the technological competence consists of six measurement items. Technological competence assesses six items which are ICT infrastructure to support MDM, organization knowledge, organization acceptance, and sufficient business and IT personnel to implement MDM (Wang & Wang, 2016). In addition, personnel expertise on MDM is also included as an item measuring the technological competence of the organization (Lin, 2006). The experts agreed that the government policy consists of five measurement items. The measurement items for government policy were adapted from different measurements discussed in various research (Awa & Ojiabo, 2016; Kuan & Chau, 2001; Lian et al., 2014; Pan & Jang, 2008). The measurement items cover the government policy by the Malaysian public sector to assess the establishment of policy in supporting data sharing, data quality management, laws to protect the organization’s interest, MDM planning in the 11th Malaysia Plan, and data security. The experts decided that the citizen demand consists of four measurement items. Three measurement items of citizen demand were adopted from Wang and Feeney's study (2016) to assess citizen demand for the integrated services, abilities to use online services, and citizen trust in silo management of services. In addition, one more item of citizen demand includes the assessment of high demand for integrated,
116 timely, and quick information through online web and mobile from the citizen ( Liang et al., 2017), The dependent variable of the conceptual model is MDM adoption. MDM adoption in this research refers to the decision of the organizations to participate as MDM data providers by sharing their master data with the MDM. The experts agreed on the six items adapted from Awa and Ojiabo (2016). The measurement items involve the intention of organizations to adopt MDM to improve their service delivery to the citizens, data quality management, operational efficiencies, and cost reduction, interagency data exchange, integration operations across agencies, and the reduction of data duplication across agencies. For the moderator variable, citizen population density is measured by three levels which are low, medium, or high (McCullough et al., 2015; Rubin et al., 2014) Based on the number of the citizen served by each local government, the experts decided to classify citizen population density of Malaysia local government into ‘low’ if less than 100,000 people, ‘medium’ if 100,000 to 3000,000 people, and ‘high’ if more than 3000,000 people served by the local government. Table 3.7 defines the measurements item for each construct after conducting the content validity by experts. The measurement items then underwent the translation process to Malay version since the potential respondents are the officers from Malaysia local government. Table 3.7 Initial measurement items for each construct Dimension Construct ID Measurement items Adapted sources Technological Relative Advantage (RA) RA1 Implementing MDM will increase the profitability of my organization through service delivery improvement (Premkumar & Roberts, 1999) RA2 Adoption of MDM will provide timely information for decisionmaking (Premkumar & Roberts, 1999)
117 Dimension Construct ID Measurement items Adapted sources RA3 Data duplication in my organization will be reduced as my organization can refer to the MDM for other related master data without having to create some new ones (VilminkoHeikkinen & Pekkola 2013) RA4 The MDM will allow my organization to cut costs in our data management operations since common master data are managed by the central repository (Premkumar & Roberts, 1999) RA5 The MDM will improve the data quality in my organization through data sharing with other public organizations (Premkumar & Roberts, 1999) Complexity (CX) CX1 Identifying master data of my organization that can be shared with MDM is difficult (Loshin, 2009) CX2 Master data of my organization need to undergo a complex data cleansing process before being shared with MDM (Loshin, 2009) CX3 Integrating MDM innovation in our current work practices will be very difficult (Premkumar & Roberts, 1999) CX4 The skills required to use MDM are too complex for our employees. (Premkumar & Roberts, 1999) DQ1 Master data in my organization are complete (DAMA UK Working Group, 2013) Quality of Master Data (DQ) DQ2 Master data in my organization are not duplicated (DAMA UK Working Group, 2013) DQ3 Master data in my organization are up-to-date (DAMA UK Working Group, 2013) DQ4 Master data in my organization are valid (DAMA UK Working Group, 2013) DQ5 Master data in my organization are accurate (DAMA UK Working Group, 2013)
118 Dimension Construct ID Measurement items Adapted sources DQ6 Master data in my organization are consistent (DAMA UK Working Group, 2013) Data Security (DS) DS1 Data exchange between my organization and central repository of the MDM requires a secured communication medium (Soliman & Janz, 2004) DS2 In the MDM repository, data is safeguarded from unauthorized changes (Soliman & Janz, 2004) DS3 In the MDM repository, sensitive master data is protected from those who should not have access to it (Soliman & Janz, 2004) DS4 MDM requires disaster management to protect data in the MDM repository from any disaster (Hristidis et al., 2010) DS5 The data exchange transactions between my organization and MDM need to have digital signature verification (Smallwood, 2014) Organizational Data Governance (DG) DG1 The stakeholder’s organization, data owner, and data stewardship for the MDM implementation will be identified (Smallwood 2014) DG2 The achievement of MDM comes from the ongoing responsibility taken (Hung et al., 2014) DG3 The MDM implementation will identify the accountability of decision making (Hung et al., 2014) DG4 My organization will follow the systematic procedure for dealing with changes caused by the implementation of MDM (Hung et al., 2014) DG5 My organization will certainly define the business cases for every initiative or application of the MDM (Hung et al., 2014) DG6 My organization will clearly define a measure to evaluate the impact of adopting MDM (Hung et al., 2014)
119 Dimension Construct ID Measurement items Adapted sources Top Management Support (TS) TS1 Top management in my organization is highly interested in using MDM (Premkumar & Roberts, 1999) TS2 Top management in my organization is aware of the benefits of MDM for the future success of the organization (Premkumar & Roberts, 1999) TS3 Top management in my organization has allocated adequate financial and human resources for the development and operation of MDM (Premkumar & Roberts, 1999) TS4 Top management has the vision to project in my organization as a leader in the promotion of MDM (Premkumar & Roberts, 1999) Technological Competence (TC) TC1 The ICT infrastructure for supporting applications integration with MDM is available in my organization (Wang & Wang, 2016) TC2 My organization contains a high level of MDM innovation knowledge (Wang & Wang, 2016) TC3 My organization contains a high level of MDM innovation acceptance (Wang & Wang, 2016) TC4 My organization is dedicated to ensuring the employees’ expertise in MDM technology (Lin, 2006) TC5 The IT expertise of the personnel in my organization is good (Wang & Wang, 2016) TC6 My organization will provide sufficient business personnel to implement MDM (Wang & Wang, 2016) TC7 My organization will provide sufficient IT personnel to implement MDM (Wang & Wang, 2016) Environmental Government Policy (GP) GP1 Government has established a policy to support data sharing among government organizations (Lian et al., 2014) GP2 Government has established a data quality management policy (M. Allen & Delton Cervo, 2015)
120 Dimension Construct ID Measurement items Adapted sources GP3 Current laws and regulations are insufficient to protect my organization’s interest (Awa & Ojiabo, 2016) GP4 MDM innovation has been established as one of the aims in the 11th Malaysia Plan (Lian et al., 2014) GP5 The government needs to establish data security policies in the operation of MDM (M. Allen & Delton Cervo, 2015) Citizen Demand (CD) CD1 Citizens demand an integrated service among local government departments from my organization (Wang & Feeney, 2016) CD2 Citizens can easily use the online services that provide services across multiple local governments units (Wang & Feeney, 2016) CD3 Silo management of services across local government authorities will lower citizen trust in local government (Wang & Feeney, 2016) CD4 Citizens have very high demand for integrated, timely, and quick information through online web and mobile (Y. Liang et al., 2017) Adoption MDM Adoption by local government (MA) MA1 My organization will adopt MDM to improve service delivery (Awa & Ojiabo, 2016) MA2 My organization will adopt MDM to improve data quality management (Awa & Ojiabo, 2016) MA3 My organization will adopt MDM to improve operational efficiencies and reduce operational costs (Awa & Ojiabo, 2016) MA4 My organization will adopt MDM to improve inter-organizational data exchange (Awa & Ojiabo, 2016) MA5 My organization will adopt MDM to reduce data duplication among government organizations (Awa & Ojiabo, 2016) MA6 My organization will adopt MDM to improve operation integration across agencies (Awa & Ojiabo, 2016)
121 Upon the final selection of measurement items to be included in the final survey using CVR calculation, CVI is computed for the whole survey test validity. CVI values were calculated for each construct by calculating the average values of CVR of the accepted or retained measurement items in the specific construct. Table 3.8 illustrates the CVI values for each construct and overall survey validity, indicating that the validity of the survey instrument was achieved at 89% of CVI. Table 3.8 Content Validity Index (CVI) of the survey instrument Construct Measurement Items Total Accepted Items (CVR >0.59) CVI Relative advantage 5 5 1.00 Complexity 4 4 1.00 Quality of master data 6 6 1.00 Data security 5 5 1.00 Data governance 6 6 0.97 Top management support 4 4 1.00 Technology competency 7 7 0.97 Government policy 5 5 0.93 Citizen demand 4 4 0.82 MDM Adoption by local government 6 6 1.00 Overall CVI 0.89 3.6.1.5 Translation and Face Validity After the content validity test was conducted, the survey was translated from English language to Malay, the native language of the potential respondents using the back-translation method. Back translation is one of the most common instrument translation methods (Maneesriwongul & Dixon, 2004). Back translation method requires at least four independent translators, two translators for forwarding translation, and two translators for back translation. Invitation email to participate in
122 the survey translation was sent to six potential translators based on their mother language, and knowledge in MDM and IS (Appendix H). Four of them were agreed to participate in the translation process. Table 3.9 depicts the translators that involved in this research. Table 3.9 Translators Translation step Translator Background Knowledge/Experience Forward translation (English – Malay) T1 MDM Practitioner, MAMPU MDM and IS T2 Lecturer in IS, Universiti Teknologi Malaysia IS Back translation (Malay – English) T3 MDM Practitioner, Private Company MDM and IS T4 Lecturer in Language Academy, Universiti Teknologi Malaysia IS The back translation method in this research was based on Sousa and Rojjanasrirat (2011). The translation started with the translation of the original survey instrument from the source language (English) to target language (Malay). This step is known as forward translation. Two independent translators were involved, their mother language is Malay, but they are also fluent in English. A first translator is a person that has a knowledge in MDM, and a second translator is a person that not have knowledge in MDM but had an experience in IS. Then, the comparison between the translated versions from both translators was done to resolve any ambiguity and discrepancy. To ensure the translation quality, the forward-translated version of the survey was translated into English back by two independent translators. This step is known back-translation. Two independent translators were involved, their mother language is English, but they are also fluent in Malay. Again, the first translator is a person that has a knowledge of MDM, and a second translator is a person that does not have knowledge in MDM but had an experience in IS. Then, the comparison between the two-source languages: 1) the back-translated version; and 2) the forward-translated version was discussed to resolve any ambiguity and discrepancy. After the forward-
123 translation and back-translation were done, the Malay version of the survey was produced. Translation confirmation from the translators is presented in Appendix I. Upon the completion of back-translation process, ten potential respondents from local government organizations in Malaysia were chosen for face validity testing in order to obtain feedback on clarity and wording of the survey. The comments from both the translators and face validity test provide a quality basis for the translated version of the survey to be used in collecting data of this study. 3.6.1.6 Pilot Testing The pilot testing is required to statistically determine the quality of the survey before the actual data collection (DeVellis, 2016). The translated survey was distributed to the respondents of the pilot testing which involve 30 personnel from local government organizations in Malaysia. 10 from City Councils, 10 from District councils, 10 from Municipal and Special Councils. Reliability analysis using Cronbach’s alpha was performed to analyze the result of the pilot test. Cronbach’s alpha refers to the function of the number of test items and the average inter-correlation among the items (Santos 1999). Table 3.10 indicates the Cronbach’s alpha values for reliability analysis result from the pilot test. The value of Cronbach’s alpha for all constructs is more than 0.7, showing that the inter-correlation among the items of the constructs is highly reliable and appropriate to be used for actual data collection (Santos 1999). Table 3.10 Reliability analysis result of pilot testing Constructs No. of items Cronbach’s Alpha Decision Relative advantage 5 0.871 Accept Complexity 4 0.846 Accept Quality of master data 6 0.874 Accept Data security 5 0.832 Accept
124 Constructs No. of items Cronbach’s Alpha Decision Data governance 6 0.952 Remove highest loading item (DG1) Top management support 4 0.805 Accept Technology competency 7 0.919 Accept Government policy 5 0.944 Accept Citizen demand 4 0.910 Accept MDM Adoption by local government 6 0.954 Remove highest loading item (MA2) The value of Cronbach’s alpha for all constructs is more than 0.7, showing that the inter-correlation among the items of the constructs is reliable. Nevertheless, DG construct shows a very high Cronbach’s alpha value (> 0.95) which may indicate a high level of item redundancy (Peterson, 1994). To ensure the reliability of the survey instrument, highest loading item DG1 from the Data Governance construct, and MA2 from MDM Adoption construct was removed from the survey items and the reliability analysis was re-run. The new Cronbach alpha for the Data Governance is (0.934) and MDM Adoption (0.936), which indicates that the item inter-correlation of the constructs is reliable. According to Peterson (1994), the acceptable range for Cronbach alpha value in applied research is between 0.7 and 0.95. After removing the DG1 and MA2 from the survey measurement, the survey instrument was finalized as shown in Appendix J. 3.6.2 Data Collection There are two (2) main steps performed in conducting data collection: 1) determine sampling size; and 2) administer the survey. The following sub-sections discuss each step in-detail.
125 3.6.2.1 Determining Sampling Size The unit of analysis of the study is the organization. The organization is an established entity that consists of a group of people to achieve the same mission, vision, strategies, and goals (Miles, 2012). The term “organization” in this research refers to the departments of local government organizations in Malaysia. Different departments in local government organization manage different master data according to the core business they handled. The adoption decision would come from the department's willingness to supply their master data to the MDM initiatives. The population of this research comprises the total of 465 departments (i.e. Information Management Department, Town Planning Department, and Business Licensing and Petty Traders Department) from 155 Malaysia local government (Johor State Government 2017, KPKT Selected Statistics 2015). The selection of these three (3) departments is based on the master data entity managed by these departments. These departments are responsible for managing master data regarding business registration and licensing. The necessary sample size to validate the research model was determined by using power analysis, which is one of the reliable methods in defining the sample size (Faul, Erdfelder, Lang, & Buchner, 2007). Using the G*Power tool, a power analysis was performed. The required sample size is computed as a function of input values for the required significance level α, the desired statistical power 12β, and the to-bedetected population effect size (Faul et al., 2007). The prior power analysis of this research uses error probability α = 0.05, the power (1- β) = 0.8, and 12 numerators (relative advantage, complexity, quality of master data, data security, data governance, top management support, technology competency, government policy, citizen demand, citizen population, moderation effect of citizen demand * citizen population, and MDM Adoption by local government). As shown in Figure 3.4, the total sample size required is at least 127. This means that minimum 127 departments are necessary to participate in this research.
126 Figure 3.4 Sampling size calculation using G Power More importantly, since this research proposes the citizen population density as a moderator on the relationship between citizen demand and MDM adoption by local government, stratified random sampling rules was applied based on the calculated total sampling size = 127. Stratified random sampling is used when the population is divided into strata or categories. In this research, the citizen population density of local government is used as a strata category to ensure reliability test in evaluating the moderation effect. First, the citizen population density frequencies in the population are determined. Second, based on calculated overall sample size = 127 departments and population size = 465 departments, a sample size of the strata was calculated using the formula (Sample size of the strata = overall sample size/population size * layer size). Layer size is a total number of strata in population. And finally, the calculation of the sample size for each strata based on the formula was presented in Table 3.11. The calculation indicates that at least 19, 42, 66 departments from high, medium, and low citizen population density are necessary to participate in the research survey.
127 Table 3.11 Sample size of the strata Citizen population density (strata) Layer size Sample size of the strata (Overall sample size/population size * layer size) High (more than 3000,000 people) 69 (23 local government x 3 departments) 19 departments Medium (100,000 – 3000,000 people) 153 (51 local government x 3 departments) 42 departments Low (less than 100,000 people) 243 (81 local government x 3 departments) 66 departments Total 465 (155 local government x 3 departments) 127 departments 3.6.2.2 Administering the Survey The survey was addressed to the 465 head of departments of Information Management, Business Licensing and Petty Traders, and Town Planning. A complete directory of all local government organizations in Malaysia was obtained from Ministry of Urban Wellbeing, Housing, and Local Government website. To administrate the survey, this research combined internet-based and paper-based approach. Even though there is skepticism about collecting data through the internet, rather than paper-based, Leung and Kember (2005) revealed that the response was roughly equal between paper-based or Internet-based data collection. Hence, to improve the response rate, this research combined both internet-based and paper-based to ease the respondents by giving them options according to their preference. The process of distributing the survey started with an Internet-based approach by sending out an email invitation to all 155 local government organizations in Malaysia to participate in the survey (Appendix K). The emails were sent with an URL link to access the survey. The online survey form was created using Google form. The cover letter with university logo was also attached to the email to increase the credibility of the research. In order to encourage the responses, respondents’ profile such as name and organization were promised to be anonymous. In addition, the
128 respondents were offered to contribute to charity organization by participating in the survey. In the first three weeks, the response rate was not encouraging wherein less than 70 respondents were involved. To follow up the non-respondents, they were contacted by phone in the following week. The follow-up calls increased the respondents to 169 respondents. Additionally, this research used a paper-based approach to retrieve the nonrespondents from the internet-based approach. Using mailed survey (Dillman & Wiley, 2007), the researcher sent the hardcopy of the survey to the target respondents by posting the survey to the respondents’ addresses. The survey was attached with a cover letter and addressed envelope with a stamp. With this approach, the responses were increased to 185 respondents. The researcher also reached the non-respondents by distributing the hardcopy survey to the potential respondents by hand in their offices. This only applied to some local government organizations in Kuala Lumpur, Putrajaya, Selangor, Negeri Sembilan, Malacca, and Pahang state. With this approach, the final number of the respondents was 227. 3.7 Phase 5: Model Validation After collecting the survey responses, the data were analysed to validate the proposed conceptual model. The main purpose of the analysis was to examine the relationship between constructs or variables in the conceptual model as proposed through the hypotheses (Zikmund et al., 2013). Table 3.12 illustrates the workflow of the data analysis steps in this research. Table 3.12 Phase 5 – Model Validation Research Phase Tasks Techniques Tools Deliverables Phase 5: Model Validation Initial Data Preparation i. Response Rate Analysis ii. Data Cleaning i. SPSS ii. Multivariate Normality Test Tool i. Response Rate Analysis (Section 5.2.1, page 166) ii. Data Cleaning (Section 5.2.2, page 167)
129 Research Phase Tasks Techniques Tools Deliverables iii. Non-Response Bias Test iv. Common Method Bias Test v. Normality Test iii. Non-Response Bias Test (Section 5.2.3, page 167) iv. Common Method Bias Test (Section 5.2.4, page 168) v. Normality Test (5.2.5, page 169) Descriptive Analysis Demographic Analysis SPSS Demographic analysis (Section 5.3.1, page 171) Inferential Analysis using PLSSEM (Model Validation) Measurement Model Analysis (PLS-SEM) SmartPLS 3.0 i. Internal Consistency Reliability (Section 5.4.1.1, page 175) ii. Convergent Validity (Section 5.4.1.2, page 176) iii. Discriminant Validity (Section 5.4.1.3, page 179) Structural Model Analysis (PLSSEM) SmartPLS 3.0 i. Collinearity Assessment (Section 5.4.2.1, page 184) ii. Path Coefficient Analysis (Section 5.4.2.2, page 185) iii. Coefficient (Section 5.4.2.3, page 188) iv. Effect Size f2 (Section 5.4.2.4, page 188) v. Blindfolding a Predictive Relevance (Section 0, page 189) vi. Effect Size q2 (Section 5.4.2.6, page 190) Moderation Effect Analysis SmartPLS 3.0 i. Moderation Effect Analysis (Section 5.4.3, page 191) ii. Simple Slope Analysis (Figure 5.10, page 194)
130 3.7.1 Initial Data Preparation Prior to statistical analysis, this research started the data analysis process with initial data preparation. The initial data preparation is an essential step in the analysis of PLS-SEM (Hair et al., 2016). Upon the completion of survey, the data collection issues that primarily required to be inspected including response rate, missing data, suspicious response pattern, non-response bias, common method bias, and data distribution was done. To address these issues, this research performed five steps during this stage, namely response rate analysis, data cleaning, non-response bias test, common method variance test, and normality test. Statistical Package for the Social Sciences (SPSS) is used for initial data preparation analysis except for normality in which additional multivariate normality tool is used and can be accessed through the URL https://webpower.psychstat.org/models/kurtosis/. Initially, response rate analysis was performed because it is important to assess the value of research findings (Baruch & Holtom, 2008). Then, the data cleaning process was conducted based on Sekaran (2016) to examine the missing data and eliminate the suspicious response pattern from the responses data. Subsequently, to check whether common method bias occurs in this research because both independent and dependent variables were collected from the same respondents, Harman’s singlefactor test and full collinearity variance inflation factors (VIF) (Kock, 2015) were applied. Common method bias is defined as the systematic error variance from the measurement with the same source or method (Podsakoff, Mackenzie, Lee, & Podsakoff, 2003). Finally, the normality test was performed to examine the distribution of the data whether they are parametric or non-parametric, using multivariate skewness and kurtosis (Ramayah et al., 2017). The description of criteria and analysis result of initial data preparation stage is discussed in Section 5.2, page 166.
131 3.7.2 Descriptive Analysis Statistical techniques consist of two main categories, namely descriptive analysis, and inferential analysis. Descriptive analysis is a statistical data analysis technique which always being performed before conducting any statistical on complex modeling. It is applied to summarize the data by describing and characterizing the data (Pagano, 2012). According to Thompson (2009), descriptive analysis is typically used to measure data frequency distribution and central tendency. A frequency distribution specifies the number of occurrences of the selected data based on specific classification. The number of occurrences may also be specified using a percentage value for each category. A frequency distribution can be illustrated by a table or graphical visualization such as line charts, pie charts, and bar charts. Meanwhile, central tendency describes the middle values of the selected data which usually represented by using mode, mean and median values. The descriptive analysis was performed against the demographic profiles of 224 valid responses data using SPSS. Demographic analysis is important to quantify the characteristic of the respondents participated in this research. Demographic characteristics involve citizen population, type of local government, departments, designation group, working experience in local government and working experience in data management. The analysis result of the demographic analysis is discussed in Section 5.3, page 171. 3.7.3 Model Validation using PLS-SEM Model validation was performed by using inferential analysis of PLS-SEM. The inferential analysis is an empirical test to describe or conclude the population based on its representative sample (Zikmund et al., 2013). According to Bryman (2008), inferential statistical are divided into three groups; univariate analysis, bivariate analysis, and multivariate analysis. The univariate analysis involves hypothesis testing of one variable at a time. The bivariate analysis involves hypothesis
132 testing of two variables at a time. While multivariate analysis involves hypothesis testing of multiple variables (three or more) or groups of variables simultaneously. The inferential statistical is based on multivariate analysis since there are various relationships between variables that formed the conceptual model. One of the techniques for analyzing multiple relationships between dependent and independent variables simultaneously is through the Structural Equation Modelling (SEM) technique (Hair et al., 2016). SEM is based on a combination of factor analysis and multiple regression analysis that seeks to analyse the relationship between multiple variables simultaneously in a systematic and comprehensive way (Straub et al., 2004). In fact, SEM enables researchers to build variables which cannot be measured directly and test theory in more flexible way (Haenlein & Kaplan, 2004). In other words, SEM can provide a comprehensive method to assess and adapt the theory so that new theory can be developed (Anderson & Gerbing, 1988). Hence, SEM is more suitable for researchers who embrace positivism with the quantitative approach (Urbach & Ahlemann, 2010). There is two dominant SEM approaches. A covariance-based approach; also known as Covariance-based-Structural Equation Modelling (CB-SEM) and composite-based approach (PLS-SEM). Both approaches differ in terms of the analysis objective, statistical assumptions and statistical result (Gefen, Straub, & Boudreau, 2000). The main goal of the CB-SEM is to measure the variables of the theory and evaluate the variables through the data so that the model fit can be achieved. Thus, CB-SEM is suitable in theory testing rather than theory development (Hair et al., 2016). However, the CB-SEM requires a large sample size and data must be normally distributed to achieve the Goodness-of-Fit index through the calculation of Chi-square, Goodness of Fit Index (GFI), Adjusted Goodness of Fit Index (AGFI), Normed Fit Index (NFI), Comparative Fit Index CFI (CFI), and Root Mean Square Error of Approximation (RMSEA). The PLS-SEM approach uses the ordinary least squares (OLS) and multiple linear regression to analyse one variable at a time (Gefen et al., 2000). This technique is important to maximize the variance of endogenous variables from exogenous
133 variables (Hair et al., 2016). Therefore, the PLS-SEM approach is more appropriate for theory exploration, model prediction and theory development (Urbach & Ahlemann, 2010). Compared to the CB-SEM, PLS-SEM does not require large sample size and data does not need to be normally distributed because PLS-SEM has a linear regression feature that is less influenced by the diversion of the normal distribution of data (Gefen et al., 2000). The motivation of choosing PLS-SEM research design over CB-SEM for this research is due to the rationale that PLS-SEM is mainly intended for causal predictive analysis between independent variables and dependent variables, whereas CB-SEM is appropriate for theory testing which highlights the shift from exploratory to confirmatory analysis (Urbach & Ahlemann, 2010). Furthermore, PLS-SEM is used to maximize the explained variance of the endogenous latent construct (Ladik & Stewart, 2008). In this research, the endogenous latent construct is MDM adoption by local government organizations in Malaysia. PLS-SEM also is a popular method to validate complex path models with latent variables and their relationships (Hair, Hollingsworth, Randolph, & Chong, 2017; Sarstedt, Ringle, & Hair, 2017). The conceptual model of this research is considered as a complex path model since it has twelve variables including independent and dependent variables. Initial reviews of SEM show that the average number of variables per model is clearly higher in PLSSEM compared to CB-SEM, which has an average of eight constructs (Sarstedt et al., 2017). In addition, PLS-SEM is appropriate for small to medium sample size, rather than a large sample size and it is robust for normal and non-normal data (Hair, Sarstedt, Hopkins, & Kuppelwieser, 2014). 3.7.3.1 Measurement Model and Structural Model Analysis PLS-SEM analysis was conducted using SmartPLS 3.0 tool. At the beginning of PLS-SEM analysis, the conceptual model was translated into PLS-SEM model consisting of two types of models, namely: 1) measurement model; and 2) structural model (Hair et al., 2016). The measurement model also known as an outer model presents the relationships between each construct or latent variable (exogenous and
134 endogenous) and its measurement items. Figure 3.5 illustrates the measurement model of the conceptual model, which highlights each construct and its measurement items. Similarly, the structural model or also known as an inner model presents the relationships among the constructs or latent variable exogenous, moderator variable and latent variable exogenous. Figure 3.6 demonstrates the structural model of the conceptual model, which highlights the relationships between constructs. From both models, the circle shapes represents the constructs to be measured; the rectangle represents the measurement items of each construct; and the arrows represent the relationships between constructs and their measurement items (Hair et al., 2016). Figure 3.5 Measurement model
135 Figure 3.6 Structural model PLS-SEM consists of two main validations which are measurement model analysis and structural model analysis. The measurement model analysis was performed to validate the measurement model or outer model by examining the internal consistency reliability, convergent validity, and discriminant validity. Whereas, the structural model analysis was conducted to validate the structural model or inner model by assessing the collinearity, path coefficient, coefficient of determination, effect size, and blindfolding and predictive relevance. Since this research also aimed to examine the moderating effect of the citizen population on the relationship between citizen demand and MDM adoption, moderation analysis was conducted to examine this effect. The criteria and analysis result of PLS-SEM analysis is discussed in Section 5.4, page 173.
136 3.8 Phase 6: Model Evaluation and Discussion The final phase of this research is to interpret the data analysis result and to discuss the findings in answering the research hypotheses. Then, this stage also proposed a set of guidelines and strategy of MDM adoption for the Malaysian public sector in order to evaluate the developed model in the Malaysian public sector. Table 3.13 illustrates the task of model evaluation and discussion phase in this research. Table 3.13 Phase 6 – Model Evaluation and Discussion Research Phase Tasks Techniques Tools Deliverables Stage 6: Model Evaluation and Discussion Hypotheses discussion Hypothesis-based discussion Heat map analysis Microsoft Word Open Heatmap Hypothesis-based discussion (Section 6.2, 6.3, 6.4, 6.5, page 197) Developing MDM adoption guidelines and strategy Questionnaire Map badges analysis Microsoft Word Batchgeo i. MDM Adoption Guidelines (Section 6.7.1, page 211) ii. MDM Adoption Strategy (Section 6.7.2, page 215) iii. MDM Adoption Guidelines and Strategy Validation (Section 6.7.3, page 218) Research Implication, Limitation and Future Work Research Objectives-based discussion Microsoft Word Conclusion (CHAPTER 7, page 221)
137 3.8.1 Hypotheses Discussion A comprehensive discussion on the data analysis results is presented in (Section 6.2, 6.3, 6.4, page 197) which is based on the research hypotheses. The result was discussed in comparison with the existing research and the context of local government organizations in Malaysia. As part of analysing the moderation effect of citizen population density on the relationship between citizen demand and MDM adoption, this research performed heat map analysis to visualize the citizen population density of local government in Malaysia. The heat map analysis was conducted using an open heat map tool that can be retrieved through the URL http://www.openheatmap.com/. The discussion on the moderation effect of citizen population density is presented in Section 6.5, page 205. 3.8.2 Developing MDM Adoption Guidelines and Strategy At this stage the research was then proposed a set of guidelines and strategy of MDM adoption for the Malaysian public sector in order to evaluate the model to solve the problem of MDM adoption by local government in the Malaysian Public Sector. The development of the MDM adoption guidelines and strategy also demonstrate how the key findings of this research could be applied to the real-phenomenon of practical world. The development of the guidelines was based on the technological, organizational, and environmental determinants affecting MDM adoption found in this research. It also incorporated related guidelines produced by MAMPU such as the circular of Open Data Implementation and Big Data Implementation in the Malaysian Public Sector. On the other hand, the moderating effect of the citizen population on the relationship between citizen demand and MDM adoption by Malaysia local government was interpreted to assist the MDM adoption strategy. Figure 3.7 describes the mapping between the determinants affecting MDM adoption found in this research and the proposed guidelines and strategy of MDM adoption to Malaysia local government.
138 Figure 3.7 Mapping between research findings, guidelines, and strategy of MDM adoption and implementation Appendix N presents the proposed MDM adoption guidelines and strategy from this research. It is also worth to mention that the guidelines and strategy were reviewed by three MDM practitioners from the Malaysian public sector to validate their appropriateness in Malaysia context. As shown in Table 3.14, two practitioners are the MDM practitioners from the central organization in Malaysia and one practitioner from local government organization was involved in the validation process. Using a questionnaire adapted from Guo, Yuan, Archer and Connelly (2011), the practitioners were asked to answer questions regarding the necessity, importance, and effectiveness of the guidelines and strategy and give feedback on the proposed guidelines and strategy. Appendix O shows the practitioners’ confirmation to involve in validating the guidelines and strategy of MDM adoption and implementation. The validation feedback from the MDM practitioners are presented in Section 6.7.3, page 218.
139 Table 3.14 Practitioners involvement in guidelines and strategy validation Practitioner Agency Roles Experience in MDM ID 10 years of experience in MDM implementation in Public Sector Public Sector ICT Expert (Information Management) Central Agency P1 15 years of experience in MDM implementation in Public Sector Public Sector ICT Expert (Information Management) Central Agency P2 18 years of experience in MDM implementation in Public Sector Local IT Manager Government Organization P3 3.8.3 Research Implications, Limitations and Recommendation for Future Work Finally, research implications, limitation, and recommendation for future work are discussed and the overall conclusion is made and presented in Chapter 7. 3.9 Chapter Summary This chapter provided a detailed description of the research methodology and research design of this research. It started with the explanation of research philosophy and followed the description of the research design which is based on PLS-SEM generic process model proposed by Urbach and Ahlemann (2010). The research design formed the research roadmap consisted of six main phases including; research problem and knowledge gaps definition, theoretical foundation, conceptual model development, instrument development and data collection, model validation and ended with model evaluation and discussion.
141 CHAPTER 4 CONCEPTUAL MODEL DEVELOPMENT 4.1 Introduction This chapter describes the conceptual model development in this research. In addition, the chapter continues with the discussion on the expert verifications on the initial conceptual model (Section 4.1.1, page 141). Subsequently, the chapter explains a conceptual model proposal (Section 4.2, page 147), research hypotheses development (Section 4.3, page 154), and operational definition (Section 4.4, page 162). 4.1.1 Expert Verifications Despite a structured and rigorous method of SLR in conducting a literature review to propose a new research model, the debate on the quality appraisal stage of SLR is still ongoing. Many scholars conclude that there is a lack of an absolute guide in selecting a sufficient quality article to be included in the review analysis (Okoli & Schabram, 2010). Hence, it is important to have further verification against the initial conceptual model through expert verifications. Expert verifications are used to verify the research model for consensus-building (Ghobadi & Daneshgar, 2010; Gobbens et al., 2010). In addition, since the data verifications for this research would involve the Malaysia local government, the expert reviews are also important to verify whether these determinants are appropriate in Malaysian context. Five experts participated in expert verifications in this research. They are from the public universities, local government organizations, and central agency. The experts were selected based on their expertise, roles in the agency, and experiences in
142 MDM and IS. The list of experts and the approach of inviting them is elaborated in Section 3.5.2, page 103. The experts were asked to rank the relevance of the selected nine determinants to be incorporated in the new conceptual model according to their point of view, based on the existing scenario of MDM adoption by local government in Malaysia. The ranking sheet used a scale of 1 to 5 (‘strongly disagree’ to ‘strongly agree’) adapted from (Normand, Mcneil, Peterson, & Palmer, 1998; Sekaran, 2016). This ranking process is important to measure the consensus among the experts whether the determinants are appropriate to be included in the conceptual model of Malaysia local government context. Likely, they can suggest any additional determinants for the conceptual model together with their opinions about the decisions that they had made. The consensus was measured using Interquartile Range (IQR) as suggested by De Vet, Brug, De Nooijer, Dijkstra and De Vries (2004). The IQR score which represents the consensus from the experts were calculated for each determinant as a rationale for whether to retain or drop the determinant from the conceptual model. Based on Gracht (2012), the determinants with IQR of 1 or less were retained, while the determinants with an IQR more than 1 were dropped. New determinants suggested by the experts were also incorporated into the conceptual model after cross-checking with the results from the SLR. Table 4.1 presents the final decision on the conceptual model based on the IQR calculation. The IQR scores for all proposed determinants were less than 1 which indicates a high level of consensus, except the determinant – cost in which the IQR score was more than 1 which indicates a low level of consensus. Experts from central government and local government (E3, E4, and E5) suggested to drop the determinant cost from the conceptual model because the implementation of MDM in Malaysian public sector mostly is fully funded by the MDM initiators which are usually central government agencies.
143 Table 4.1 Expert verificati No. Determinants of MDM Adoption in Local Government E1 E2 E3 E 1 Relative Advantage 5 5 5 5 2 Complexity 5 5 5 5 3 Security 5 5 5 5 4 Cost 4 4 1 1 5 Governance 4 5 5 5 6 Top Management Support 5 5 5 5 7 Technological Competence 5 5 5 5 8 Policy and Regulation 5 5 5 4 9 Citizen Demand 5 4 5 5 Additional suggestion from experts None None - Change ‘Governance’ to ‘Data Governance’ - Change ‘Policy and Regulation’ to ‘Government Policy’ - ‘Quality data’ as a technolog determina - Add a mo effect of cit population the relation between cit demand and adoption
ion on initial conceptual model E4 E5 Median Q1 Q3 IQR (Q3- Q1) Level of Consensus (Decision) 5 5 5 5 5 0 High 5 5 5 5 5 0 High 5 5 5 5 5 0 High 1 1 1 1 4 3 Low 5 5 5 5 5 0 High 5 5 5 5 5 0 High 5 5 5 5 5 0 High 4 5 5 5 5 0 High 5 5 5 5 5 0 High of master new ical ant oderation tizen density to nship tizen d MDM - ‘Quality of master data’ as a new technological determinant - Add a moderation effect of citizen population density to the relationship between citizen demand and MDM adoption Not Applicable
144 One of the experts (E3) suggested that data governance plays an important role in the adoption of MDM by local government in Malaysia. He advised to change the determinant ‘Governance’ to ‘Data Governance’, so that it would be more reflective on the MDM domain. As data governance in MDM implementation would clearly define the roles and responsibilities of data ownership, data stewardship, and the steering committee of the MDM initiative, it was perceived to encourage the local government in Malaysia to adopt MDM. He also advised to change the determinant ‘Policy and Regulation’ to ‘Government Policy’. This is due to policy and regulation is two different aspects. The policy is the statement or the standard of the fundamental decision by the people such as elected parliament, government, or commission, while regulation is the decided task of setting down the practical rules in compliance with policies (Fransman, 2010). Taking these statements together, both policy and regulation could lead to a very broad scope and it is recommended to specifically use the term ‘Government Policy’ in the conceptual model. Government policy refers to the policies that are outlined by the Malaysian Government. In addition, the experts (E4 and E5) from local government suggested to include quality of master data in the organization as a technological determinant that affects local government adoption of MDM. They stated that one of the barriers which is preventing the local government to provide their master data to an MDM central platform is the poor quality of master data they have in their organization. With the poor quality of master data – e.g. incomplete, redundant, invalid, obsolete, inaccurate, and inconsistent – the local government may face difficulties during the data cleansing process before they share the data with the MDM. Any discrepancy in the data may affect their reputation as a data owner. Especially in Malaysia, according to the Data Quality Index (World Economics, 2017), Malaysia has achieved 79.4% of the overall data quality indicators, which is lower than in other developing Asia countries such as Singapore, Israel, Bahrain, and Qatar with 93.3%, 90.9%, 82.5%, and 81.0%, respectively. Even though the data quality assessment only involved the evaluation of Gross Domestic Product (GDP) data, this index may represent the scenario of data quality in the Malaysia local government context.
145 Moreover, E4 and E5 also suggested the moderation effect of citizen population density to the relationship between citizen demand and MDM adoption. They argue that citizen demand towards integrated government online services through MDM adoption is influenced by the citizen population density of local government. The citizen population density of a local government could possibly determine by the government type (i.e. city council has more than 500,000 citizens, municipal council has between 100,000 to 500,000 citizens, district council, and special council has less than 100,000 citizens) (KPKT, 2017a). However, in some cases, the categorization of local government is influenced by political and tourism purposes. As such, Langkawi Municipal Council (approximately 92,784 people), Pasir Gudang Municipal Council (approximately 46,571 people), and Putrajaya Corporation (approximately 68,361 people) are being considered as municipal councils even though the citizen population is low (less than 100,000 people). Hence, the experts suggested that citizen population density can be classified into three levels which are ‘low’ if less than 100,000 people, ‘medium’ if 100,000 to 300,000 people, and ‘high’ if more than 300,000 people served by the local government. From the expert verifications, the initial conceptual model was modified based on the experts’ suggestions as recommended in Figure 4.1. As it can be seen, eight determinants were agreed to be retained with some changes on the term used in the conceptual model (i.e. relative advantage, complexity, data security, data governance, top management support, technological competence, government policy, and citizen demand). One determinant i.e. cost was dropped from the model because most of the MDM implementations in the Malaysian public sector are fully funded by the MDM initiators, not by the local government organizations. One additional determinant i.e. quality of master data was added in the technological dimension due to its appropriateness in the context of Malaysia local government and it is a crucial determinant in the MDM domain. Similarly, the moderation effect of citizen population density to the relationship between citizen demand and MDM adoption was also proposed by the expert to the conceptual model.
146 Figure 4.1 Modification of initial conc
eptual model based on expert verifications
147 4.2 Conceptual Model All suggestions from the experts were cross-checked on previous studies of IS theories and literature before adding or modifying the initial conceptual model to support the theoretical foundation of the conceptual model. The reference sources of all proposed determinants and the relationships between them are shown in Table 4.2. As it can be seen, all suggestions from the experts have strong theoretical foundations hence it is significant to be included in the conceptual model. Table 4.2 Theoretical foundation of the Conceptual Model Relationship Source Domain Technological Relative advantage -> MDM adoption by local government (Rogers, 1995) Theory of DOI (Dedrick et al., 2014; Gao & Lee, 2017; Kamal, Hackney, & Ali, 2013; Kamal, Hackney, & Sarwar, 2013; Liang et al., 2017; Sharif et al., 2015) Local Governments (Alharbi, 2016; Bonnet, 2013; Otto, 2012; Vilminko-Heikkinen & Pekkola, 2013) MDM Complexity -> MDM adoption by local government (Rogers, 1995) Theory of DOI (Lagrandeur & Moreau, 2014) Local Governments (Loshin, 2009) MDM Quality of Master Data -> MDM adoption by local government (Wang & Strong, 1996) Theory of Data Quality Framework (Kwon, Lee, & Shin, 2014) Organization (Silvola et al., 2011) MDM Data Security (Ali et al., 2016; Lagrandeur & Moreau, 2014; Rubin et al., 2014) Local Governments
148 Relationship Source Domain -> MDM adoption by local government (Piedrabuena et al., 2015; VilminkoHeikkinen & Pekkola, 2013) MDM Organizational Data Governance -> MDM adoption by local government (Ali et al., 2016; Liang et al., 2017; McCullough et al., 2015; Velleman et al., 2015; Wang & Feeney, 2016) Local Governments (Alharbi, 2016; Bonnet, 2013; Dreibelbis et al., 2008; Haug et al., 2013; Loshin, 2009; Otto & Schmidt, 2010; Smallwood, 2014; Smith & McKeen, 2008; VilminkoHeikkinen & Pekkola, 2013) MDM Top Management Support -> MDM adoption by local government (Dedrick et al., 2014; Kamal, Hackney, & Ali, 2013; Kamal, Hackney, & Sarwar, 2013; Liang et al., 2017; Rubin et al., 2014; Seigler, 2017; Velleman et al., 2015) Local Governments (Silvola et al., 2011; Vilminko-Heikkinen & Pekkola, 2013) MDM Top Management Support -> Data Governance (Tjan, 2001) Theory of Fit-Viability framework (Elliott et al., 2013; Liang et al., 2007) Organization Top Management Support -> Technological Competence (Tjan, 2001) Theory of Fit-Viability framework (Haque & Anwar, 2012; Hoffman, Hoelscher, & Sherif, 2005; Liang et al., 2007) Organization Technological Competence -> (Ali et al., 2016; Dedrick et al., 2014; Gao & Lee, 2017; Kamal, Hackney, & Ali, 2013; Kamal, Hackney, & Sarwar, 2013; Lagrandeur & Moreau, 2014; Liang et al., Local Governments
149 Relationship Source Domain MDM adoption by local government 2017; McCullough et al., 2015; Norris & Reddick, 2013; Rubin et al., 2014; Seigler, 2017; Sharif et al., 2015; Velleman et al., 2015; Wang & Feeney, 2016) (Silvola et al., 2011) (Bonnet, 2013) (Haug et al., 2013) MDM Environmental Government Policy -> MDM adoption by local government (Ali et al., 2016; Dedrick et al., 2014; Jans et al., 2016; Kamal, Hackney, & Sarwar, 2013; Liang et al., 2017; Rubin et al., 2014; Sharif et al., 2015; Velleman et al., 2015; Welch et al., 2016) Local Governments (Dreibelbis et al., 2008; Haug et al., 2013; Otto et al., 2012; Spruit & Pietzka, 2014) MDM Citizen Demand -> MDM adoption by local government (Dedrick et al., 2014; Liang et al., 2017; Sharif et al., 2015; Velleman et al., 2015; Wang & Feeney, 2016) Local Governments (Dreibelbis et al., 2008; Otto et al., 2012) MDM Citizen Population Density * Citizen Demand (Gao & Lee, 2017; McCullough et al., 2015; Rubin et al., 2014) Local Governments 4.2.1 Technological dimension The technological dimension describes the characteristic of the innovation relevant to the organization (Baker, 2012; Tornatzky & Fleischer, 1990). Figure 4.2 illustrates four proposed technological determinants of the conceptual model. It presents two determinants from the theory of Diffusions of Innovation which are relative advantage and complexity. According to (Rogers, 1995), relative advantage refers to the extent in which the innovation could increase ROI, reduce operating costs,
150 resolve current problems and receive various benefits. Complexity on the other hand, refers to the extent of organization difficulty to understand, implement and use the innovation. Other determinants proposed in the technological dimension are quality of master data and data security. Quality of master data refers to the extent of completeness, uniqueness, timeliness, validity, accuracy, and consistency of master data at the local government organizations. Data Security refers to the extent where the innovation could preserve information’s confidentiality. Figure 4.2 Technological dimension in the conceptual model 4.2.2 Organizational dimension The organizational dimension refers to the resources characteristic (Baker, 2012; Tornatzky & Fleischer, 1990) and linking structure of the personnel of the organization (Krishnan et al., 2017; Smallwood, 2014). As it is seen in Figure 4.3, the organizational dimension of the conceptual model proposed three determinants which are data governance, top management support, and technological competence. Data governance is a subset of information governance, which involves processes and controls of the information at the data level (Smallwood, 2014). Top management support refers to the extent in which the management supports the adoption of IT innovation (Premkumar & Roberts, 1999). Technological Competence refers to the ICT infrastructure readiness, knowledge, skills, experience and a sufficient number of personnel to implement the technology (Wang & Wang, 2016).
151 In addition, within the organizational dimension, it is also posited that the top management support will influence the data governance and technological competence. This relationship is a mirror to the viability context in Fit-Viability Model, FVM proposed by Tjan (2001) (Section 2.3.3, page 50). The FVM described that the viability or organizational readiness is higher if the organizational factor such as top management support is higher. According to (Wang & Wang, 2016), technological competence and data governance are categorized as organizational readiness which refers “to the extent in which an organization perceives its capability to integrate technology-related infrastructure, professionals, expertise, and skills for innovative technology implementation”. Figure 4.3 Organizational dimension in the conceptual model 4.2.3 Environmental dimension The environmental dimension refers to the arena in which the organization conducts its business (Baker, 2012; Tornatzky & Fleischer, 1990). The environmental dimension consists of two determinants which are government policy and citizen
152 demand as illustrated in Figure 4.4. Government policy evaluates the existence of fundamental policies or standards to adopt the MDM by local government. Citizen demand towards the IT innovation in relation to these businesses would influence the IT adoption. As illustrated, one moderator variable was introduced in the conceptual model, namely citizen population density. It is posited that citizen population density moderates the relationship between citizen demand and MDM Adoption by local government. Figure 4.4 Environment dimension in the conceptual model Mutually, the technological, organizational, and environmental proposal propositioning a new MDM adoption model for Malaysia local government as described in Figure 4.5. The conceptual model presents the MDM adoption by the local government as a dependent variable and nine technological, organizational, and environmental potential determinants as independent variables. One variable i.e. citizen population density is introduced as a moderator to the relationship between citizen demand and MDM adoption by local government. The explanation of each relationship or also known as a hypothesis is discussed in the following section.
153 Figure 4.5
Conceptual model
154 4.3 Research Hypotheses Development The hypothesis is the statement of prediction towards the relationship between variables of the conceptual model. Hypotheses development is the next logical step after the formulation of the theory or conceptual model (Sekaran, 2016). Subsequently, twelve research hypotheses were formulated to explain the relationships among variables in the conceptual model of this research. 4.3.1 Relative Advantage Relative advantage measures the extent in which technology innovation could increase ROI, reduce operating costs, resolve current problems and receive various benefits (Premkumar & Roberts, 1999). Theory of Diffusion of Innovation by Rogers (1995) defines relative advantage as a crucial characteristic of the IT innovation that influences the adoption decisions by the organization. In the context of local government, previous studies found that relative advantage plays a key role in organizational-level IT adoption, such as e-government (Kamal, Hackney, & Sarwar, 2013), enterprise application integration (Kamal, Hackney, & Ali, 2013), and social media (Sharif et al., 2015). As identified by Otto (2012), MDM intents to provide an authoritative source of high-quality master data, lowering operational cost and complexity through standards, and support of business intelligence and data integration. Similarly, Vilminko-Heikkinen and Pekkola (2013) argued that MDM would lower the operational cost by avoiding master data to be maintained in more than one location which promote an effective data management by streamlining data management scope of works. In addition, Alharbi (2016) and Bonnet (2013) in their study maintained that clean and accurate consolidated data in MDM will encourage organizations to get a decision-enabling data analysis when they adopting MDM. In view of this, the first hypothesis is: H1: Relative Advantage has a positive effect on the MDM adoption by local government
155 4.3.2 Complexity Complexity is one of the technological characteristic proposed by Diffusion of Innovation theory (Rogers, 1995), which refers to the extent of organization’s difficulty to understand, implement and use the innovation. Since complexity typically has a negative effect on the adoption of innovation, it is seen as a mean to inhibit adoption of new innovation (Premkumar & Roberts, 1999; Tornatzky & Fleischer, 1990). Lagrandeur and Moreau (2014) stated that complexity influences organization’s decision to adopt the e-services by municipal governments. Complexity involves the effort an organization makes to learn and use new technology. The easier an organization perceives a new technology to be learned and used, the less complex it perceives it to be learned and used, and vice versa. According to Loshin (2009), introducing MDM into an organization is a complex process and requires numerous steps and viewpoints which may lead to the reluctant of organizations to adopt MDM. Thus, it is proposed that: H2: Complexity has a negative effect on the MDM adoption by local government 4.3.3 Quality of Master Data Quality of master data in this research measures the degree of completeness, uniqueness, timeliness, validity, accuracy, and consistency of master data at the organization (DAMA UK Working Group, 2013). Low quality of master data in the organization may delay the adoption process of MDM. Silvola et al. (2011) in their study mentioned that master data in different formats and unreliable data has hindered the consolidation process at the beginning of MDM implementation due to the complexity of cleansing process. This leads to slow and difficult MDM adoption by the organization. The data quality measurement involves the assessment of completeness (the degree of completeness of master data in the sources which it is measured by comparing the presence of non-blank values against a hundred per cent), uniqueness (the extent of uniqueness of master data in the sources which it is measured by analysing the number of things as assessed in the 'real world' compared to the
156 number of entities in the master data set), timeliness (the extent of up-to-date records of master data in the sources), validity (the degree of master data at the sources that conform to the syntax i.e. format and type), accuracy (the degree in which master data at the sources correctly describes the real-world object or event being described) and consistency (the degree of similarity of one or more representatives of master data entities in the sources). This leads to the third hypothesis: H3: The quality of master data has a positive effect on the MDM adoption by local government 4.3.4 Data Security Data security refers to the degree in which inter-organizational information systems could preserve data confidentiality, integrity and availability (Soliman & Janz, 2004). Local government organizations concern about the security and privacy of their organization information when they intend to adopt particular innovation, such as cloud-computing (Ali et al., 2016), e-services (Lagrandeur & Moreau, 2014), social media (Rubin et al., 2014), and Enterprise application integration (Kamal, Hackney, & Ali, 2013). As explained by Vilminko-Heikkinen and Pekkola (2013), data security and privacy are the first general issue to be considered at each step of MDM development. Furthermore, a successful MDM have to comply with data protection regulations which may hinder the extensiveness reuse of information in a government context (Piedrabuena et al., 2015). And therefore, it is posited that: H4: Data security has a positive effect on the MDM adoption by local government 4.3.5 Data Governance Data governance includes a formal process of roles and responsibilities corresponding to the levels of authority and accountability (Smallwood, 2014). It also
157 analyses the outcome of the technology implementation, encourages persistent responsibilities, defines the systematic procedure and business case, and evaluates the impact of the technology. A good governance will fasten the process of adopting innovation by local government, such as the adoption of cloud computing (Ali et al., 2016), e-services (Wang & Feeney, 2016), Web Accessibility Standard (Velleman et al., 2015), and Electronic Health Record (McCullough et al., 2015). MDM which comprises numerous design decisions and multiple parties in the implementation requires the identification of roles and responsibilities in managing the shared master data or also known as data governance (Otto & Schmidt, 2010). Data governance is a subset of information governance, which involves processes and control of the information at the data level (Smallwood, 2014). Lack of clarity of roles in relation to data creation, use, and maintenance may lead to MDM failure. Hence, we proposed that: H5: Data governance has a positive effect on the MDM adoption by local government 4.3.6 Top Management Support Top management support refers to the degree in top management of the organization creates a supportive environment and provides adequate resources for the adoption of IT innovation (Premkumar & Roberts, 1999). Top management support includes the extent in which senior managers understand the perceived benefits of the technology, clear about the vision about applying the technology in the organization, and allocate sufficient fund and resources toward the technology implementation (Dong et al., 2009). This factor is as an important determinant of the local government to adopt social media (Rubin et al., 2014; Seigler, 2017), web accessibility standard (Velleman et al., 2015), Smart Grid Technologies (Dedrick et al., 2014), Enterprise application integration (Kamal, Hackney, & Ali, 2013), and e-Government (Kamal, Hackney, & Sarwar, 2013). With regards to MDM, Silvola et al. (2011) argued that the vision and the support from the managerial level are the success criteria for the effective MDM implementation. In addition, Vilminko-Heikkinen and Pekkola (2013)
158 highlight that management strategy is an important determinant influencing MDM development. Therefore, the next hypothesis is proposed as follows: H6: Top Management Support has a positive effect on the MDM adoption by local government 4.3.7 Top Management Support and Data Governance A strong top management support is the first key step which governs data effectively (Smallwood, 2014). Similarly, as good governance requires crossfunctional collaboration from multiple stakeholders, top management support from data provider organizations is needed to drive the establishment of effective data governance by participating and giving continuous commitment based on the defined accountability. Top management support will ensure the effectiveness of data governance through their participants and responsibility in the governance committee, while weak support from top management and lack of budget allocation leads to a poor coordination between IT and business people (Elliott et al., 2013), hence it is a major challenge to effective enterprise data management. Thus, the next hypothesis is as follows: H7: Top Management Support has a positive effect on the Data Governance 4.3.8 Top Management Support and Technological Competence Fit-Viability theory outlines that top management support as an organizational factor is one of the success factors to the viability of the IT adoption (Tjan, 2001). Viability according to Fit-Viability framework refers to the readiness of ICT infrastructure and literacy of project team member towards technology (Liang et al., 2007). To implement MDM, technical knowledge and skills are required. Thisincludes data quality, data integration, metadata management, and master data repository. The
159 organization needs competent personnel to adopt the technology. Higher support from top management will increase the technological competence of personnel through budget allocation for technical training and change management programs. Top management support has turned up as the main supporter of a positive change in the professional behaviour of workers (Hoffman et al., 2005). Top management support in term of providing funds and necessary support to enhance the capabilities of employees in creating, sharing, storing and dissemination of knowledge significantly increase the application of knowledge among banking sector employees (Haque and Anwar 2012). Thus, we propose that: H8: Top Management Support has a positive effect on the Technological Competence 4.3.9 Technological Competence Technological Competence refers to the extent of organization’s capability which includes IT infrastructure and human resources availability in terms of expertise, skills, sufficient number of personnel to adopt and implement IT innovation (Lin, 2006; Wang & Wang, 2016). In local government context, this factor encourages the adoption of Social Media (Seigler, 2017), Cloud Computing (Ali et al., 2016), eServices (Lagrandeur & Moreau, 2014; Wang & Feeney, 2016), Web Accessibility Standard (Velleman et al., 2015), Electronic Health Record (McCullough et al., 2015), Smart Grid Technologies (Dedrick et al., 2014), Enterprise application integration (Kamal, Hackney, & Ali, 2013), and e-Government (Kamal, Hackney, & Sarwar, 2013). Sufficient resources play a key role in adopting IT innovation by local government such as Social Media (Rubin et al., 2014; Seigler, 2017), Electronic Health Record (McCullough et al., 2015), and e-Government (Kamal, Hackney, & Sarwar, 2013; Norris & Reddick, 2013). In adopting MDM, the organization is recommended to gather knowledgeable and experienced personnel from diverse business units to ensure in-depth understanding and adequate MDM implementation (Bonnet, 2013; Duff, 2005; Haug et al., 2013; Silvola et al., 2011; Spruit & Pietzka, 2014). Thus, the next hypothesis is as follows: