This archive report was first published on 20 June 2020.
On October 26, 2017, Kenya held a repeat presidential election, which was closely watched by the international community. In a bid to increase transparency, the Independent Electoral and Boundaries Commission (IEBC) released a dataset of voting records from the election. The data, which covers around 40,000 polling stations, reveals interesting insights into voter identification methods.
The dataset, which was automatically recorded in the logs of the Kenya Integrated Election Management System (KIEMS) kit, shows that Mombasa County had the largest percentage of voters identified biometrically (88 per cent), while Nyamira County had the lowest percentage number at 57 per cent. This means that 12 per cent of voters in Mombasa County and 43 per cent of the voters in Nyamira County were allowed to vote using 'other means' that were not biometric.
According to the data, Nyamira County had the largest percentage of voters not identified biometrically, followed closely by Kericho at (40 per cent), Wajir at (38 per cent), Nyeri at (38 per cent), and Tharaka Nithi at (38 per cent) respectively. The IEBC has been urged to investigate and explain why a standard electronic voter identification kit (EVID) would behave differently across these different counties.
One of the reasons for the use of non-biometric voter identification methods is that some voters may have unreadable fingerprints, perhaps arising from the nature of their work or accidents prior to voting day. The election law was subsequently amended, and IEBC allowed to clear voters outside the biometric identification framework.
There were two main categories of non-biometric voter identification: the 'Alphanumeric' approach and the 'Document Search' category. In the 'Alphanumeric' approach, the IEBC clerk would receive the voter's national ID card and type the corresponding ID number into the EVID Kit. Approximately 1.6 million voters were identified through this approach. The 'Document Search' category was used for voters whose biometric features were originally not captured. Approximately 400,000 voters were cleared through this approach.
Non-biometric clearance has its weaknesses, as it can be abused by having absentee voters cleared. To mitigate this, additional procedural requirements were completed upon clearing voters using non-biometric means, including the completion and signing of a form 32A by agents as one of the supporting documents.
The dataset reveals that around two million voters required some form of supporting document. This includes the 1.6 million 'Alphanumeric' voters and 400,000 'Document search' voters. The data also shows that Kiambu county registered the highest number of voters cleared to vote using other means, with around 200,000 such voters.
According to the data scientist who analyzed the dataset, the ability to derive meaning from huge datasets is crucial in informing conversations and driving policy actions. The data visualizations presented in this article have contributed towards this direction.
Mr. Walubengo is a lecturer at Multimedia University of Kenya, Faculty of Computing and IT. Email: [email protected], Twitter: @Jwalu
