*5.1. Characteristics of Respondents*

The sample respondents included 157 (78%) males and 44 (22%) females (Figure 3). Despite their low numbers, the presence of women land administration professionals in the woreda land administration offices would help the policy reform move towards gender-sensitive land tenure security. About 188 (87%) of the respondents were between 21 and 40 years of age. In addition, 82% and 18% had completed their bachelor's degrees and diplomas, respectively. This also indicates that the woreda land administration offices are filled with relatively young and degreed land administration professionals. This would likely foster innovative technology acceptance in the rural land administration sector and facilitate the establishment of the NRLAIS at woreda level. The range of disciplines the university graduates had studied were very broad, including surveying (17%), ICT and computer science (16%), agriculture (14%), natural resource management (14%), geography (12%), land administration (8%), economics (7%), and others (13%). The diversity of the disciplines would also reflect the multi-disciplinary nature of the land administration domain. However, the number of existing land administration professionals with land law and economics backgrounds appeared low.

In terms of work experience, respondents were asked how long they had been working for their respective woreda land administration offices. The experience levels ranged from less than a year (8%), to between one and seven years (70%), to over seven years (22%). The result revealed that about 61% of the respondents had worked for over 5 years, while 25% had worked for between 3 and 5 years. Only 24% of the respondents had worked less than 3 years in their respective woreda land administration offices. During data collection, on average, 83% of the land administration expert positions were filled. Despite frequent staff turnover reported during the key informant interviews as a key challenge for NRLAIS deployment, the survey result revealed a substantial level of staff retention in the woreda land administration offices. However, this does not mean that the reported land administration experts' turnover did not affect NRLAIS roll-out and activation.

In terms of system use, 67.2% of the respondents had NRLAIS use experience between six months and one year. In addition, 26.2% of the respondents had used NRLAIS for over one year and less than two years. Only 1.5% of the respondents had used NRLAIS for over three years, which probably indicates respondents from the pilot woreda of NRLAIS (Figure 4). This system use experience revealed that all respondents have had adequate familiarity with NRLAIS functional and operational issues.

Respondents were also asked how many minutes or hours per day they spent working on NRLAIS to discharge their service delivery related to land transaction management. As presented in Figure 5, 47% of the respondents spent between six to eight hours. Moreover, 31% of the respondents spent between four to six hours of office hours using NRLAIS to process land transaction management and service delivery. This indicates that about 78% of respondents use NRLAIS for over half of typical office hours. In addition, this shows that NRLAIS is being used as a source and maintains land record information at woreda land administration offices.

Similarly, the respondents were asked how many times a day on average they log into the system. The question measures the frequency of system login as a proxy indicator of access security awareness and rule compliance. The survey shows that about 45% of respondents answered that they log in over ten times per workday. About 30% of respondents log in between six and ten times per workday. Only 2% of respondents log in once per day and process the land transaction management tasks assigned to them on the system. This result seems congruent with the average number of land transactions processed (5 to 13 per day) at the woreda land administration offices. This, in turn, implied relatively good compliance with standard producers and rules by the woreda

land administration offices. However, the 2% of responses indicating only one log in per day seemed to reflect a misunderstanding of the question, as NRLAIS has a session time-out functionality.

**Figure 4.** Experience of the respondents' actual usage of the NRLAIS in their daily official business discharge.

**Figure 5.** Time spent to process land transaction management and service delivery using NRLAIS.

According to the technical specification of NRLAIS, user-specific roles and user administration have been defined through the business processes. NRLAIS also supports user role definition, assignment, auditing, and reporting with separate management trees according to the administrative structure (federal, region, zone, and woreda) and within the same hierarchy of the woreda land administration offices [74]. To this end, respondents were asked which access privileges they were assigned as internal system users or operators. As presented in Figure 6, the respondents answered that about 44% held an expert role, 31% an officer role, 21% a supervisor role, and 3% a system administrator specific role. However, the 3% responses most likely misunderstand the question, since system administration specific roles are assigned at federal, regional, and sometimes zonal or mobile IT support teams only.

**Figure 6.** Users' specific assigned role of respondents.

Similarly, according to the architecture design of NRLAIS, the woreda is the primary level at which the system functions. The technical requirement also specifies five subsystems at the woreda level, which affect only parts of the system while processing transactions management, including web information, cadastral maintenance, property registration, document management, and process subsystems [28]. The processing subsystem handles the main operations to register rights, while the cadaster maintenance subsystem handles the management of spatial features (parcels, maps, points, boundaries, etc.). Both the document management subsystem and the processing subsystem have been customized for specific rural land registration processes. Respondents were asked which subsystems they have been using the most while processing land transaction management and service delivery in priority orders. The result revealed that all respondents were used the web information subsystem, as this is the user interface to log in to access the other subsystems. This is followed by the property registration subsystem (48%), cadastral maintenance subsystem (35%), and process subsystem and document management subsystems (17%).

#### *5.2. Validity and Reliability*

The first test conducted in this study was testing the validity and reliability of the outer model. The outer model testing was performed through a process of algorithm iteration, a parameter of measurement model that includes convergence validity, discriminant validity, composite reliability, and Cronbach's alpha. Validity and reliability ensure that the multiple indications of each latent variable in the measurement model converge to measure a single construct and hence develop legitimacy, defined as the level to which things used to measure can calculate the idea they meant to quantify [75]. All items used to measure the construct should pile essentially to their constructs rather than different builds. As for the component analysis, it ensures that items are designated to their constructs, as they express high loading on them that stands out from several constructs [76]. The measurement model assessment of vertical collinearity is presented in Table 1. This shows the subjective independence of every indicator on its latent variable using cross-loading criteria.

The individual item reliability was evaluated by examining the loading and crossloadings of indicators on their respective construct. According to Fornell and Larcker's criteria [77], a reliability score of Cronbach alpha 0.6 is considered minimally acceptable, with 0.70 preferred (50% of the explained variance). The theory also recommends that an indicator loading having a value of less than 0.40 should be removed from the model. Hence, this study found three indicator items with less than or equal to 0.4 outer loading. As per the rules, the indicator items removed from the model include INQU4 (format), PEOU4 (clear and understandable), and SYQU4 (risk of losing data).


**Table 1.** Indicator item cross loading.

The composite reliability index and the average variance extracted (AVE) were applied to assess the internal consistency and convergent validity [77]. According to the rule, the square root of the AVE of a particular construct should also be greater than its correlation with other constructs. Generally, the AVE should be higher than 0.5. Table 2 shows the internal consistency of each construct. In the measurement model, the study used Cronbach's alpha and composite reliability (CR) to test the reliability of the constructs. The study found that all the CRs were higher than the recommended value of 0.700, ranging from 0.822 to 0.887. The Cronbach's alpha of each construct exceeded the recommended 0.700 threshold, which is 0.720 to 0.831 in the current study. Hence, convergence validity was acceptable, because the average variance extracted (AVE) was over 0.500.

Discriminant validity concerns the uniqueness of a construct, whether the phenomenon captured by a construct is unique and not reflected in the model by the other construct [75]. The subjective independence can help reduce the presence of multicollinearity amongst the latent variables, denoting that the average variance extracted (AVE) of a latent variable should be higher than the squared correlations between the latent variable and all other variables [77]. Discriminant validity was assessed by the Fornell–Larcker criterion [77]. Table 3 shows that the square-root of AVE for the construct in the diagonal

was greater than the inner-construct correlation, which ranges from 0.732 to 0.816. The test result of the current study may therefore imply the strong reliability of all the items.


**Table 2.** Reliability and Validity.

**Table 3.** Fornell–Larcker Criterion. Note: Value in diagonal represent the Square-root of AVE.

