“And do not cloak (and confuse) the truth with falsehood. Do not suppress the truth knowingly.” We are not to lie, even if this might win us an advantage. Quran(2:42)
An article by Dr Abusaleh Shariff and Mr Khalid Saifullah in EPW on 23 May 2018, made allegations of bias against Muslims and Dalits by wantonly deleting their names from the electoral rolls. With my analyses of the data on Chief Electoral Officer, Karnataka (CEO-KA), website, in a meeting at the office of CEO-KA on 07 Nov 2018, I showed to Dr Shariff that there was no bias against Muslim voters. He had then agreed with me. I thereafter met Dr Shariff and his colleague Mr Zameer Pasha twice more and discussed the matter. I also wrote about my findings to them, allaying the fear of targeted exclusion of voters from specific communities.
While there was no response from them, similar inconsistent arguments about the exclusion of Muslim and Dalit voters are being published in several national and international journals and newspapers. See some links at scroll.in, Frontline, Caravan, Foreign Policy, and Vartha Bharathi. We find a dozen other such links. There are also several YouTube links to talks by Mr Saifullah emphasising these points.
The latest one is an article in Aljazeera with a tagline “Activists say 40 million Muslims and 30 million Dalits not on electoral rolls, raising fears of targeted suppression”, authored by Ashish Malhotra and published on 30 Apr 2019.
Applying the logic used by the ‘researchers’ quoted by Aljazeera, we can also infer that more than 267 million non-Muslim voters are excluded from the electoral rolls of the country. Census reports of 2011 state that 41.1% of Indians are below 18. The current population of India is 136 crores This makes about 80 crore citizens eligible to be in the electoral rolls. Taking out the 30 crore excluded voters, we could have 50 crore records in the electoral rolls. It is widely reported that we have 90 crore voters. How do we account for this bloat of 40 crores? If we also confuse the counts with the missing women voters reported in the article, 50% of the entries in the electoral rolls would be illegal. Any takers? Heads at ECI could reel. None will roll. Our arguments sink due to their own absurdity.
“Torture the data enough and it will confess to anything” – Ronald Coase.
Aljazeera’s article starts by reporting from ‘hearsay’ reports on the mass deletion of voters from the electoral rolls. The first part of the article covers the exclusion of voters due to controversial citizenship issues. I am not qualified to comment on this.
I had sent rejoinders to the claims to EPW, Times of India, Economic Times and the author of the article in https://foreignpolicy.com. My rejoinders were not published by any one of them.
Aljazeera states, “Overall, it is hard to ascertain just how many people are not on India’s voting rolls. Research by two leading psephologists has suggested that as many as 28 million women are missing from the electoral rolls.”
CEO Karnataka statement to the press announced, “There were 956 female voters for every 1,000 male voters in 2014 and 972 in 2018. Now, it stands at 976.” As per census report of 2011, Karnataka has 973 females for every 1,000 males. At least in Karnataka, women are over-represented in the electoral rolls. The authors say “as many as 28 million women are missing from the electoral rolls,” but do not mention the count of missing men from the electoral rolls. It is a misleading incomplete statement.
A naive reader is prone to infer a bias against women while the statistics quoted by CEO-KA say otherwise.
It is reported that Mr Saifullah runs a software company doing research on electoral rolls. However, his findings do not seem to be based on facts. He told Al Jazeera that “he believes around 65 million of these missing voters are women.” The basis for his belief is not clear. 65 million women, 40 million Muslims, 30 million Dalits … we lose 100 million voters! (40 million Muslims + 30 million Dalits + 65 million women – 35 million Muslim and Dalit women = 100 million )
Then, Aljazeera makes a damning statement, “Among the 120 million people Missing Voters suggests are not included on the rolls, Saifullah further estimates that around 40 million of them, like Saleh and Imran, are Muslims while 30 million are Dalits, the former untouchables.”
There are several flaws in this argument, based on the data patterns in the electoral rolls of Bangalore with 18% of Karnataka voters. “Saifullah’s estimates are based on discrepancies he says he found between the number of single households in census data and the election commission’s data. The numbers include those who may not have registered to vote in addition to the names that may have been deleted.”
- ECI allows homeless people to enrol as voters. Such voters would not have house addresses.
- House address is formed with the house number and section address, which is like a street address. Most of the section addresses are not detailed enough to find the location of the houses. In Bangalore, on an average, a part has 1,060 voters. 28% of them have only one section each. Putting about 1,060 voters in a single street address shows that the data is grossly incomplete and inadequate. 15% of the parts have 2 sections each, and 10% have 3 sections each.
- In many cases, people from the same house are shown in different houses in the electoral rolls. E.g., husband and wife living in a house are shown in different houses, and even in different parts.
- People are shown in wrong houses in a large number of cases.
- Per cent of single-voter Muslim houses is about equal to the per cent of Muslim voters. Therefore if Saifullah concludes that 40 million Muslim voters are excluded based on the count of single-voter Muslim households, 267 million non-Muslims also would have been excluded from the rolls.
We cannot trust the measurements taken using inaccurate and un-calibrated equipment. Allegations based on such measurements would mislead the readers. Then, there are repeated calls for action based on such ‘research’ inferences.
The statement “There are also concerns that India’s minorities, mainly Muslims and Dalits, were the chief targets of voter suppression,” in the article cannot be substantiated. The following points highlight the flaws/problem with the approach:
- Whereas we can identify Islamic names, it is not possible to identify the religion, caste and/or communities of most of the other voters. I consider it mischief to state that Dalits are being excluded from the electoral rolls. On 07 Nov 2018, Dr Shariff had agreed that it was only a guess.
- It is technically non-trivial to re-create voter records from the currently published English versions of electoral rolls. It is a slow process and demands high computing resources.
- It is extremely difficult to re-create the records from electoral rolls published in Indic scripts. About 90% of the electoral rolls are published only in Indic scripts.
Mr Saifullah’s talks, widely circulated on YouTube, claiming that his company has re-created the voter records for the entire country. In light of the foregoing, this sounds like a tall claim. It would be good to know how his team did this and to see a demonstration. If it is done with the common good in mind, the technology should be shared.
From the electoral rolls of Bangalore, published on26 Mar 2019, the counts of additions and deletions between Oct 2018 and Mar 2019 are as follows: –
|Total Analysed||Islamic Names||Islamic %|
The electoral rolls of Bangalore have about 13% Muslim voters, but only 9.44% of the deletions bear Islamic names while 11.36% of the additions bear Islamic names. If at all there is a bias in the electoral rolls of Bangalore,, it is in favour of the Muslims and not against them. The concern expressed, “There are also concerns that India’s minorities, mainly Muslims and Dalits, were the chief targets of voter suppression” is a concern as it appears to be false propaganda based on our analysis.
We cannot comment on Saifullah’s allegation that no verification is done when people file Form 7. We cannot state so unless it is proven. If the allegation is true, it is very serious, requiring immediate investigation and action by the ECI.
The article quotes Srinivas Kodali, “When you’re deleted from a roll, you never know which year you were deleted in? Was it 2015 or 2017? There is no way to know.” In all the versions of the electoral rolls published by the CEOs on their websites except the ones published as mother rolls in September/October, the supplements show the details of additions, deletions, and modifications. If a person looks at the version prior to the mother rolls, it is easy to find the date of deletion of the voters. The CEOs also publish the details of claims and objections from which we can know the additions and deletions in response to the forms submitted by citizens. It is not as opaque as made out in the article. Thus this is an unsubstantiated charge.
The same article by Kodali states “They have issues with their algorithms and their systems which they don’t want to accept … They want to keep using the same system or process forever”. Whereas the system does need improvement in many areas, there is no doubt that there have been several improvements over the years. Many data errors and inconsistencies which were common a few years ago have been corrected. During the past year, various State-level Electoral Roll Management Systems have been integrated into a single central system named ERO Net, which we hope to bring several benefits.
Despite several improvements over the last few years, the electoral rolls still have major quality issues related to data quality and data completeness. Also, some process guidelines from the ECI are regularly flouted on the field. However, I have not seen any malpractice due to political influence or any bias based on sex, religion, caste, or community.