Electoral rolls are open for claims and objections during continuous updation period when the ERMS (Electoral Roll Management System) is not under maintenance. Claims are in Form-6, application for inclusion in the rolls by people whose names do not appear in the rolls. Objections are in Form-7, requesting deletion of names, Form 8/8A, seeking corrections to voter data. The forms can be submitted online (followed with hard copies of proofs) or in hard copies to EROs.
Periodically, integrated versions of the rolls are printed. Else, the rolls have a main part and supplements with Additions, Deletions and Modifications lists. A document may have several supplements till it is next integrated. The last integrated version of rolls of Bangalore city was published in December 2012. Supplements were added in Jan 2013 and April 2013.
While modifying, a record is marked by prefixing ‘#’ to its serial number in the main part of the document. Modified version is printed in Modifications List Supplements.
In the rolls of Bangalore city, 1,67,341 voter records were modified after the version of Dec 2012 till 20 April 2013. With the high volume of work, some operator errors are expected. Some glaring problems and inconsistencies as a result of modifications expose very poor quality of the software that allows such callous errors. Even the basic principle of transaction closure is not honoured by the software. It is incorrect to blame the process or operators for the errors which the application can block.
For long I have been reporting that modifications lead to data corruption. However, the rolls continue to be polluted while adding as well as while correcting records.
During the corrections since December 2012, the process of modifications has contributed to large scale corruption of data leading to deletion of about 6,000 innocent voters. The sequences appear to be as in the following example:
Serial 129 of AC1510022.pdf has record of Sujith. She finds that her name, age, and father’s name require corrections and files Form-8. Her record gets corrected and appears in the Modifications List in the document.
The correction does not end, like it should happen in a regular transaction. In AC1510027.pdf, Sherin Mary Anthoni’s record at serial 286 gets modified with data matching that of Sujitha Ganesan.
Sherin Mary suffers an identity theft, loses her right to vote, and Sujitha gets two votes, in two different booths. Online search at http://ceokarnataka.kar.nic.in/SearchWithEpicNo_New.aspx# for voter IDs UZJ0073361 and UZJ4660107 give two records for Sujitha. We do not find any record for Sherin Mary.
This is the genral pattern. However, in Sujitha’s case, the riddle does not end. We do not know how and when Sherin Mary’s record got to the rolls. AC1510027.pdf published in Nov 2012 had Sujith’s name at serial 286. In the rolls published in December 2012, Sherin Mary’s name appears, gets modified and disappears. A poltergeist.
There are more than 6,000 cases like this where the record before and after corrections are not of the same person.
1,718 records appear in Modified List, but their originals are not in the main part of the rolls. Thus, we do not know which record is modified. Many serial numbers of modified records are higher than the last serial number in the list. E.g., AC1760176.pdf states 1434 as the ending serial number at page 1. We find up to serials 1433 and total number of voters in the summary table at the end of document is 1428. Record in Modified List have serial numbers 3892 and 5970.
In addition, 176 records marked with ‘#’ do not have matching modified versions. Modification was intended, but not done.
Introducing new Errors
As per EC guidelines, names should not have prefixes. Since June 2012 I have requested the CEO a few times to remove prefixes like Dr., Prof., Mrs., Late, etc. Instead of acting on the request, many records are ‘corrected’ to introduce the disallowed prefixes to names.
Some modifications have corrupted the records further. E.g., XTE1076116 Lingappa’s sex was ‘Male’ and appeared correct. With the changes, his (her?) sex has become ‘Female’ and his father’s name has become ‘Late.Narayanappa’ from ‘Doddanarayanappa’
Summary of Corrections
Using fuzzy-token-matching and ignoring cases (upper/lower), scores in the range of 0 to 100 are given to matches between pre- and post-modified versions of the records. As an effect of the modifications alone, in two months about 18,000 records are effectively deleted and an equal number of duplicate records are created.
- Respect the guidelines and schema given by EC. Data validation should automatically correct where appropriate, reject illegal entries, and warn when it cannot decide.
- Software should ensure transaction closure. In an activity like modification of records with more than one atomic step, either it should complete all the steps or roll back to the previous state. This will avoid imbalances in records.
- Validate every entry with other related data in the repository. E.g., if the part has recorded a number as ‘last serial number,’ update the entry when a serial number larger than that is inserted. Ensure that there are no gaps in the serial numbers.
- During modifications, software should compare the data and warn if the new data does not resemble that of old record. If the match is lower than a threshold, do not allow the corrections.
- In all cases, if data is entered despite warning, mark the record for validation before approval by the ERO. Leave an audit trail of such approvals.
- Make the software vendor and operators accountable for errors.