This supplementary report is an extension of the #TeamPachedu Technical Analysis, dated the 16th of July, providing further evidence to support the issues raised in that report. A number of anomalies were explored in the previous technical report as the issue of ghost entries was explored. This study, however, transcends beyond being exploratory and confirmatory, but leans more towards being an expository. Our findings broadly confirm and provide evidence that the 2018 voters’ roll, released under the guise of being biometric-based and thus infallible, does in fact contain ghost entries.
In this report, we identify and explore 3 main strategies that have been used to create ghost entries in the voters’ roll, and the numbers, albeit being relatively low, are based on what we can confidently prove. However, the actual total is beyond the figures below, which were identified under very strict anomaly detection models.
Strategy 1: New identities were created from existing/once-existing identities. The total number of cases where an ID was reassigned to the same name or different names with different dates of birth was 128,096.
Strategy 2: The systematic change of ID suffixes. In Zimbabwe, ID suffixes are issued based on ones’ place or origin, or rather, the ancestral district. These suffixes are inherited from the father’s ID Number. It would be very unusual for a person to have the suffix changed between 2013 and 2018. A total of 30,802 cases were changed in the 2018 voters’ roll from the 2013 roll, and within the 2018 voter’s roll per se, 2180 more cases with suffixes that share the same ID exist. In virtually all the cases where the suffix was changed, or where there were two IDs with different suffixes, only one has a voting history, and the other is a first occurrence regardless of age.
Strategy 3: Assignment of a new ID number, in part or in full, to individuals registered in 2013 with the same name, same surname and same date of birth. Arguments that names and surnames could be identical were factored in during our extraction model parameters. The 10,182 records obtained were those that the extraction models deemed to be statistically impossible. The total of those initially flagged is actually more than 40,000, but, herein, we present the main anomalies alone.
Based on these methods, we now have a list of names that we are 95% confident that if checked by contesting parties in their respective wards and constituencies, or otherwise, the people with those identities will never be found. Over an above, the full list from each of the above tests shall be shared in the supporting files archive.
Source: Team Pachedu
Access supporting files (17 MB Zip Folder)
[/vc_column_text][vc_row_inner column_margin=”default” top_padding=”20″ text_align=”left”][vc_column_inner column_padding=”no-extra-padding” column_padding_position=”all” background_color_opacity=”1″ width=”1/1″ column_border_width=”none” column_border_style=”solid”][text-with-icon icon_type=”font_icon” icon=”icon-file-text” color=”Extra-Color-3″]Download PDF (1.7MB PDF)[/text-with-icon][/vc_column_inner][/vc_row_inner][/vc_column][/vc_row]