Integrity of the Electoral Rolls

Integrity of the Electoral Rolls

Cdr PG Bhat (Retd.), 16 Mar 2020

Abbreviations

BLA	Booth Level Agent (of the political parties)
BLO	Booth Level Officer
BLV	Booth Level Volunteer (citizen volunteer)
CEO	Chief Electoral Officer – one per state/UT
DEO	District Electoral Officer
ECI	Election Commission of India
EPIC	Elector’s Photo Identity Card
ERMS	Electoral Roll Management System
ERO	Electoral Registration Officer – one per constituency
EVP	Elector Verification Program
NVSP	National Voter Services Portal
SSR	Special Summary Revision

Note: A voter is referred in masculine gender for brevity. Please read he as she/he and his as her/his.

1. 480 Assembly Constituencies (of the 4,121) from 14 different states/UTs are  required to publish electoral rolls in English as per  the Hand Book for EROs. We have analysed every version of the electoral rolls of the 28 constituencies of Bangalore published since 2010 and have occasionally analysed other electoral rolls published in English. The observations in this document are mainly based on the electoral rolls of Bangalore though most of the observations apply to others as well.  Various counts quoted in this document refer to the electoral rolls of Bangalore.
2. Since the management of electoral rolls is now centralised with ERO-Net application of ECI, improvements to the system will benefit the entire nation because all the states and UTs now use this system. 

3. Data Accessibility

CEOs publish new versions of electoral rolls on their websites at least twice a year and more often in election years. ECI expects the voters to check every new version of electoral rolls to raise claims and objections. EROs delete or change  voter data without due diligence and without notifications to the affected party. Every time a new version of the electoral rolls is published, it is important for a voter to check if he is a bonafide voter and if his data in the rolls are correct. He can do so by reading the electoral rolls on CEO  website or by querying by EPIC number or by name at NVSP. Both the approaches have some difficulties.

1. The electoral rolls are published as image PDF files with CAPTCHA protection. One electoral roll maps to one part in an assembly constituency, which has one booth in almost all cases. To look up the names in the electoral rolls, the voter should know his constituency and the part number. Very few voters know this data. The part number of voters change with delimitations. Voter serial number in the part could change with delimitation and with the draft version published in Sep/Oct. Adding to the confusion, often some names are in parts other than where they should be due to wrong addresses in the electoral rolls.
2. The electoral rolls were being published as text PDF files until changed by ECI letter No.485/Comp/ERO-Net/2017 dated 04 Jan 2018 directing that “Electoral Roll should be published on Website in image PDF only… No other format of Electoral Rolls be made available in open domain… The access to view this image of Electoral Roll be strictly provided through CAPTCHA…”
3. The electoral rolls are public documents but not open data. Being image documents, we cannot search text except by scrolling the records one by one. Searching voter records in the electoral rolls is impractical for a common man. EROs and their teams also face the same problem.
4. Extracting voter records from the image files for analysis is a non-trivial task. It demands better resources than the ubiquitous professional/home PCs. Also, the results have large number of typos because the process involves optical character recognition (OCR).
   • Why should a citizen analyse the electoral rolls?
   • During the past decade, some lapses in the management of electoral rolls and several quality issues in the data were reported by the citizens resulting in corrections to some of them by the CEOs/ECI.
   • Patterns injustice and errors (discussed later) cannot be found without analysis of the electoral rolls. Though such analyses by citizens have helped in identifying grave errors in the past, since end 2017, the ECI has been trying to block access to voter data. This, however, is not a deterrent to the people/organisations with good resources. Published on the Internet, it is accessible to the world. People with resources have no real restrictions but the citizens who need to know the data find it difficult to access.
5. If one knows the EPIC number, searching at NVSP is easy. Search by name is not effective because it does not have fuzzy search feature. Unless one enters the name exactly as it is entered in the rolls, the record would not be found. With many typos and no standard for entering the names, often the search results in false negative cases. Similar is the case with voter helpline app which creates chaos close to the elections.
6. We regularly find a few unreadable voter records in the scanned PDF file as well as on query at NVSP.
7. Another Example of Data Denial. We understand that in Sep/Oct 2018, a team led by Sri Wajahat Habibullah discussed a few lapses in the system, citing examples from the electoral rolls of Karnataka found by Dr Abusaleh Shariff. One of the points was that 18% of applications for registration were rejected with a reason that the applicants were not citizens of India. The data was extracted from the status of claims and objections published on CEO-KA website. This prompted the CEO to investigate the matter and discover a software bug that reported wrong reasons for rejection of applications. Whereas the software error was corrected, the CEO also stopped giving the reasons for rejection thus blocking future possibilities of error discovery by the citizens. 

4. Unsynchronised Data

For the past 5 years, often the data at NVSP has been out of synch with the electoral rolls.

E.g., CEO-KA published the latest version of electoral rolls on 07 Feb 2020. When we search for the voters at NVSP website, we do not find some voters who are in the published lists. The pattern observed thus far is that these are mostly (but not all) the last records in the List of Additions. E.g., EPIC numbers XHL5809777 and UZJ7908676 found in AC170  Part 131 and AC151 Part 002 respectively are not found with search at NVSP.

This was communicated to the ECI and CEO-KA on 19 Feb 2020 and reminders were sent on 24 Feb,  08 Mar, and 16 Mar. The error is not corrected and the authorities have not answered the emails.

When a person does not find his name in the electoral rolls, he is expected to register again. When he does so, because of poor quality checks in the system, duplicate records are created.

5. Wrongful Deletion of Voters 

1. Citizens of several states have complained in the past  about lakhs of valid voters being deleted from the electoral rolls. An Election Commissioner and some CEOs had accepted such lapses in the press and in public meetings.
2. Deletion for convenience. As an example, CEO-KA had deleted about 13.5 lakh out of 65 lakh voters of Bangalore in 2012 because the records did not have voter photographs though the EPICs had. Reason entered for the deletions was ‘Shifted Residence.’ The voters were not intimated of the deletions. CEO-KA had tried to cover his lapses by illegally deleting valid voters rather than correcting the errors until ordered to do so by the  High Court, Karnataka. Ref: WP 38844-47/2012 (GM-RES-PIL).
3. Deletions due to wrong addresses. Many voter addresses are wrong in the electoral rolls. When the BLO would not find the voter in the recorded address during inspections, he is expected to recommend deletion of such voters after due diligence. Though wrong house addresses were to be corrected during the EVP/SSR between Aug and Dec 2019, we do not find these corrections in the electoral rolls published subsequently.
4. Deletion by over-writing. During correction of data in response to requests from the citizens, at times wrong records are updated. This results in deletion of some valid voters and duplicate records for the other. Though such occurrences have significantly reduced, the deleted records have not been restored. Changes have not been made to the data entry software to stop such occurrences.
5. Deletions during delimitation. In several states, voter records (even more than a lakh records) are lost during delimitation.
6. Though ECI rules allow restoration of wrongly deleted voters, it is seldom done. The citizen is not informed of the deletion. He is expected to find out that his name is missing from the electoral rolls by inspecting the electoral rolls and then apply for registration. The officials ask such people to state in the application that the person was never registered as a voter, which is a lie.

Claims and Objection Status

The ERO is required to intimate the progress in the processing of various applications from the citizens for inclusion, deletion, and corrections to voter records. Neither the mandated acknowledgement is given to the citizens when they apply nor are they intimated of the progress. Earlier, CEO-KA used to publish the status of applications on his website periodically. The data now being published is not complete and does not adhere to one standard format. The UI is different for different constituencies. Access of data is made more difficult and confusing.

• Duplicate Entries

The electoral rolls are replete with duplicate entries. With every new version of the electoral rolls, more duplicate entries are added.

1. ERO-Net is ineffective in detecting duplicate records.

Consider the following two cases in the electoral rolls published in Dec 2019.

Case 1: AC170, Part 93
Sec	Sl	House	EPIC	Name	Sex	Age	Relative	Reln
1	78	86	XHL5783931	AKSHATHA L	F	22	LOKESH M S	F
1	79	86	XHL5781752	AKSHATHA L	F	22	LOKESH M S	F
Case 2: AC170, Part 6
1	138	17/1	XHL5662143	AKASHATHA B S	F	23	SRINIVAS B	F
1	139	17/1	XHL5477609	AKSHATHA B S	F	23	SRINIVAS B	F

The first case would be reported as Demographically Similar Entry (DSE) – suspected duplicate entry – by ERO-Net because the voter names and relative names exactly match. The second case is not reported as DSE because the names do not match exactly.

• We discover much larger number of suspected duplicate records by matching the (voter-name + relative-name) phonetically. Subsequent to a discussion with the Dy EC in 2013, we group the suspected duplicate records as follows:
  • Rank 1: Same EPIC numbers. These are indisputable errors.
  • Rank 2: Same house (part, section, and house number). Photo comparison by EROs has shown that more than 90% of these suspects have matching photographs and ground checks prove them to be duplicated records.
  • Rank3: Same part but in different houses. EROs find about 80% of these suspects to be duplicated records.
  • Rank 4: Across parts. About 20% of them are duplicated records.

The following table gives a sample comparison of the suspected duplicates found by ERO-Net and by our method in October 2019:

Constituency	Within Part	Across Parts
ERO-Net	Phonetic	ERO-Net	Phonetic
AC168	61	1,096	211	37,638
AX170	198	1,456	485	26,439
AC173	102	1,866	131	15,356

• A Better and Proven Approach.

Using metaphones and fuzzy-matching approaches, we can identify similar voter-name + relative-name sets. Then, photographs of the voters within the sets can be compared using image comparison software. If the photographs match, then those voters can be physically verified and records deleted if found to be duplicated. We can add additional filters and refine the application based on experience of using this first solution.

Voter photographs are not available to the public and hence we cannot compare the photographs. We implemented image comparison feature using a set of about 1,000 voter records with photographs provided by ECI in 2013. In majority of the cases where the names matched the photographs also matched. The software was installed at ECI and proven. In 2014, CEO-TN took a copy of the software from us and found it to be useful. He had also shared a small set of voter records with photographs with us.

Why the approach is not adopted by ECI despite being proven is a mystery.

We have been sending the lists of suspected duplicate entries to CEO-KA for the past 8 years and to the EROs since last year. There has not been any action or even acknowledgement of the emails so far.

• Improve Data Entry Software. Software can easily suspect and warn the operator when a duplicate entry is being made. If the operator over-rides the warning and enters the record, it can be flagged for verification by the ERO.

• Shifted/Dead People in the Electoral Rolls
• People who have shifted residence continue to be in the electoral rolls of the old addresses. ECI rules allow a person to remain an elector in his ‘permanent address.’ The rules also require deletion of voters not found in the registered address. This has created a contradiction and confusion. Citizens tend to register in more than one place though this would require a false declaration.  Data entry software does not detect duplicate entries.

• Large number of dead people are still in the electoral rolls. In addition to expecting the relatives to request deletion of the such entries, the CEOs can gather data from the Birth and Death Registrations. Part of such data is available digitally and would make updating the electoral rolls easier. E.g., see  

•  The following table shows the % people at various old-age ranges in Bangalore as per census-2011 data and voters in those age-ranges. Though we cannot conclude that voters list has dead voters only from these numbers, this is an indicator that rather than being happy, the ECI and CEO should check if the voter records of the older people are all genuine.

Age Range – Years	Census	Voters
80 to 90	0.71%	2.13%
91 to 99	0.18%	0.22%
100 and above	0.05%	0.06%

• By ECI rules, only a voter or a family member can request for deletion of a voter record. When a family shifts residence, the members often do not request to delete their names from the electoral rolls. Also, the grieving members of a family do not take initiative to get the name of a deceased off the rolls. RWAs, offices of building societies, and citizens could be encouraged to recommend deletion of names of non-bonafide voters in their area/building. Deletion would happen after due diligence  as per the current rules.

• Inadequate Data
• Sections. A part (one electoral roll) is divided into several sections. A section is defined as an area with geographical boundaries and is like a street address. Most of the parts are not organised into sections. E.g., of the 8,514 parts in Bangalore, 2,340 parts have only one section each and 1,273 have 2 sections each. We have average 1080 voters per part in Bangalore. One street address for as many voters is not logical.

• Section addresses. Section addresses are mostly inadequate to find the location as also pointed out by the Assistant Post Master General of Karnataka in a meeting with CEO-KA.

• House addresses. Though ECI has prescribed house numbering standards, it is not yet implemented. ECI has directed that notional house numbers be given where a dwelling place does not have a door number. However, a large number of house numbers are entered as 0 or left blank.  House address is formed by joining house numbers with section addresses. With inadequate/wrong house numbers and section addresses, many voters cannot be easily located. Wrong addresses also result in illegal deletion of voter records.

• Wrong Houses. In many cases, house addresses are wrong. Voters of a family living udner  one roof are shown in different houses, often in different parts. In some cases, houses are shown in a different address and in different parts. This type of wrong data creates confusion and also to disenfranchisement. Though this was to be corrected during the EVP in 2019, all the errors are still in the electoral rolls.

1.   Wrong Data

Wrong names, sex, age,  and relationship are common in the electoral rolls. There are several cases where  a husband is shown as father and father-in-law as husband. Some voter records show males having husbands. Whereas it is the responsibility of the citizens to get wrong entries corrected, ECI software can include features to detect logical errors and initiate steps to correct.

1.   Unaccounted EPIC

Whereas the EPIC serves as an identity card in many places and also is a proof of citizenship, they are not accounted. Some examples:

1. When a voter changes his constituency, he is expected to register again rather than only changing his address. He is issued with a new EPIC without having to return the old one.

• If a voter record is deleted illegally, the voter has to apply for registration even when the illegal deletion is proven. He is asked to state that he was never registered. The person gets new EPIC without returning the old one.

In Hyderabad, a lady’s voter record was deleted without reason. When she took up the case with the CEO, she was told to register again stating that she had never registered as a voter. She did as directed and then self-surrendered for making false statement. The police refused to file the case.

1.   Roles and Responsibilities

EROs/AEROs/BLOs cannot handle the responsibilities entrusted to them. They could be given better support with good software tools. Much of the work expected out of them can be done at higher levels with better computing facilities and access to data. E.g., a BLO is expected to submit various statistical reports. He may not be even a matriculate and works as BLO in addition to his regular government job. His compensation is Rs. 6,000 per year.

Most of the BLOs exist only on paper. Those who do exist don’t know what they are expected to do. Having tasked them with what they cannot do, those tasks are not taken up by the higher authorities. The system is designed for inefficiency. 

1.   No Response to Citizens’ Feedback and Offer of Help

The ECI and CEOs seldom respond or acknowledge feedback from the citizens. We have been offering help to the organisation for the past ten years.

The ECI launched an ambitious-on-paper program named EVP/SSR, conducted between Aug and Dec 2019 with an objective of improving the quality of electoral rolls. It ended with no effort to meet any of the goals notified in the policy letters. Orders that could not be implemented were passed. Lower level officials were expected to do without support and without necessary tools.

We had a few meetings with CEO-KA in this period and offered several helps. There was no encouragement. Despite that, we provided the following data to the 28 EROs of Bangalore:

1. Suspected duplicate records in their constituency, ranking 1 to 4 as discussed earlier.
2. Voter records organised by houses for each part. ECI had directed the EROs to provide such lists to BLOs but the EROs had no capability or system support to do this.

The CEO was copied in all the communication with the EROs. Neither the EROs used the data nor the CEO asked them to use it.

1.   Citizen Experience
2. ECI has published the guidelines for the contents in CEO websites. Ideally, the websites of all the CEOs of the country should have similar user experience, look and feel. They are diverse and confusing. Standardisation of the GUI in these sites will help the citizens and also will reduce the cost of maintenance of these sites.

• The sites have many broken/dead links. Data on these websites are not reliable. E.g., most of the BLOs could not be contacted using the details given on the websites. Janaagraha had made a report on this issue after surveying the in various parts of the country. 

• Close to elections, the websites become painfully slow and unusable.

• CEOs’ organisation often does not honour the prescribed processes. Escalation process does not work because the authorities do not respond to emails and letters from the citizens.

______________________________________________

Notes for CCE

Whereas I have received support and encouragement in my work, the analysis itself was done solely by me. I can provide supporting data for all my findings. Adhering to style guides, ‘we’ has been used in the document above instead of ‘I’.

Although I have studied Electoral Rolls of 486 constituencies at different times, my comments are based on systematic and regular study of the ERMS of Bangalore for a decade. The issues confronted in my persistent efforts in the ERMS of Bangalore are so fundamental as to reasonably conclude that they have universal applicability to the ERMS of the country as a whole.

My approach to the issues has been to help cleanse ERMS but it has not succeeded because the CEO, Karnataka, and the ECI have not appreciated the issues and their implication. They have been mostly unwilling to consider solutions offered on the ground.

A few of the major issues, each of which is elaborated with reference to Bangalore ERMS in the report above:

1. There are lakhs of illegal voter records. A simple comparison with the Census data and the Electoral roll adjusted for growth in population since the Census 2011 will reveal the enormity of illegal entries.

• The Electoral Rolls should be such as to facilitate easy accessibility to the voters. The published roll in PDF image form looks formidable and not meant to facilitate reference even to a computer-savvy person. As a sidelight, the rolls are available with unauthorized persons for them to sell the information. It is easy to blame the voter to have allegedly given a long name (many Indians have 3 or 4 components in their name such as father’s name, village name, given name, family/ surname and the like) but that does not solve the issue.

• Like-wise, the house address poses problems for lack of uniformity or standards in recording the addresses.

• Deletions from the voter’s list are galore and done cavalierly. At the end of each voting day, we find thousands complaining how their names were not found in the rolls. It is easy to blame the voters but the fault also lies in made-difficult accessibility to the voters’ list.

• It looks as though the ECI wants the verification by the voter made as difficult as possible so that NON-CITIZENS do not appear in the list. This approach stems from the attitude that to exclude even 1% non-citizen stealthily entering the list, other 99% genuine votes ought to undergo a punitive process.

• Owing to the above, the voters whose names are not easily traced are asked to file an application for registration which are often not updated or recorded differently in some part of the voter list resulting in duplicate entries.

All of the above problems could be reasonably and satisfactorily solved with software. I have been offering help to implement such solutions and have demonstrated its efficacy to some officers of ECI/CEO. It is difficult to avoid the impression that some CEOs do not care and some may be helpless due to lack of support from ECI.

CCE gives us hope.