Chapter 4: Data Capture, Transmission and Management
Kenya
The choice to use mobile technology for data collection was based on the existing extensive use of mobile technology in Kenya, wide network coverage, reduced price of owning mobile devices, and improved ICT literacy levels among other reasons.
Android tablets and phones were chosen due to their affordability and availability of platforms for mobile application development tools. Mobile technologies were used for cartographic mapping, pilot tested and used in the main census enumeration.
Two local universities premised on their proven experience in successful delivery of tablets were used to assemble mobile devices and installed census software and related census files.
The collected data was sent to the central server by enumerators after completion of every household or several households depending on the availability of internet connectivity. Coordinators, information and communications technology (ICT), and content supervisors supported by county and sub-county census committee members coordinated and supervised the exercise. In the event of a tablet’s failure to send the data to the server, the enumerators shared their data via Bluetooth with their respective content supervisors, who eventually sent it to the server.
All enumerators sent their data to the server at the headquarters daily and whenever there was no internet, they would send data to the supervisor via Bluetooth that was configured between the supervisor’s and enumerator’s devices. The supervisors would run pre-selected reports and review the data to check for its completeness before uploading the data to the server. A robust server was thus required to handle the magnitude of traffic and hits that were expected during the enumeration process. At the back end, the authorised personnel were pulling the data from the server to evaluate its quality and inform the respective field personnel in case data was inconsistent. In addition, they were also preparing monitoring reports.
The data processing team embarked on the process of assessing the census enumeration coverage immediately after data collection. This process involved the harmonisation of Enumeration Area (EA) Geo-codes that had been shared during tablets’ assembly and those used during enumeration. There was need for confirmation that all the EAs created during the mapping exercise were accounted for (with or without data). The next step was extraction of data from tablets whose data did not reach the server by close of enumeration. This was followed by checking and confirming whether all the EAs had data both from pre-enumeration listing of households and enumeration data. A comparison of listed households and actual returns, and accounting for the variances was done. In addition, a comparison of mapped households, listed households and actual enumerated households was also done and any discrepancies addressed.
The next step was to check for data consistency at household and individual levels. Basic edit specifications were developed to check the inconsistencies. The purpose of editing was to ensure that the final dataset did not include training data (data that was collected during training of census personnel) as well as ineligible individuals (empty records). In addition, the editing process took the following into account: households with missing household heads; individual records without age, sex, and relationship for conventional households; individual records without sex and age for non-conventional households; and duplicate records for structure, household numbers, and enumerator codes. Subsequently, tabulation plans were developed and used to generate summary tables for basic and analytical reports. Data in the generated tables was compared with existing data sets to assess the consistency of enumeration coverage and the outputs were validated by key stakeholders and thereafter launched.
Census data editing: Kenya utilised the handbook on Population and Housing Census (PHC) Editing to come up with the editing specifications for the different modules in the census; Census data editing Team: The editing team was composed of data processors, subject matter specialists (demographic and non-demographic), computer programmers, supervisors and managers; Kenya National Bureau of Statistics (KNBS) has the inhouse capacity to program the edit specifications and run them to generate clean data files; We also utilised experienced CSPro data processors (former KNBS staff) who helped in the programming of editing/ imputation codes. Analytical Reports: Subject matter specialists drawn from the 2019 Kenya PHC drafted the reports and then shared with the stakeholders for review; KNBS received technical assistance (TA) support through the UNFPA ESARO office to come up with Population Projections report; Looking forward to receiving TA to come up with Geospatial Dissemination System. Data Quality checks: An edit specification document should be prepared by the subject matter specialists and shared before pilot is undertaken; Hard checks in CAPI should be enforced for majority of the questions to reduce content errors.
Data security was achieved through several methods including encryption, secure file transfer and authentication. The system used for data transmission from the mobile devices to the census servers employed industry-leading encryption mechanisms to ensure that data was always encrypted and secure while in transit. In addition, secure sockets layer (SSL) technology and strong encryption keys were implemented at the point of collection and transmission to the server.
To ensure the security of the mobile devices including the application and data, the following was implemented; the mobile devices were running on a tested up-to-date android operating system, there was a logon PIN/Password required to protect against unauthorised access to the device, Mobile Device Management (MDM) software was installed to track the devices in case of theft. The MDM had an application locking software to protect installed apps from being uninstalled and introduction of unauthorised apps. Lastly, the devices were protected with a screen protector and flip cover to provide shock protection in case of falls. To ensure data security, a Private Access Point Name (APN) was configured to ensure that the SIM Cards could only communicate to the census servers.
Sierra Leone
Sierra Leone procured the EA Pad software 4.0 license and satellite imagery with recent footprint of all structures and other features. Data was downloaded via Google cloud from the mapping assistants in the field to the respective District Offices using onscreen digitisation of delineated Enumeration Areas using the ArcGIS Mapping software tool.
Each technological instrument intended for use in census data collection activity was available during the preparation period, and fully tested with other related applications.
The local mobile companies provided the internet services at an agreed fee, and this played a vital role working with servers at Statistics Sierra Leone headquarters to aid transfer of data directly from the field to main offices.
The mapping PAD Application (a Data acquisition package built by Milsat Technologies) did a direct export of data to CSV format and the Headquarter offices downloaded the data through a cloud server system.
For data security reasons, only accredited functionaries were allowed to use the app, the mapping PAD would periodically run system authentication on every phone the app is installed. The mapping PAD remote submission process utilised a top-level security cloud system to scan and safely deliver uploaded tasks to GIS portal. The GIS portal could only be opened by registered and verified GIS Officers.
Botswana
Data transfer to and from the servers was done through mobile data from sim cards loaded with different data packages depending on the role of end users for all the tablets or use of WI-FI.
Statistics Botswana used fort iGATE firewall as the first layer of security to protect internal network from external untrusted traffic. They made sure that the antivirus for the organisation is installed on all the servers including that of the Census for data protection.
Overall, the 2022 PHC had four end users with authorised access to tablets, laptops, and desktops. In addition, each IT personnel used their passwords to log in to the servers.