Kolkata: Before the third phase of the Election Commission’s Special Intensive Revision (SIR) of voter rolls, Delhi had 1,45,10,299 registered voters. By August 17, Booth Level Officers (BLOs) working house to house had collected forms from 97,53,577 of them. The rest, 47,56,722 names, did not appear on the draft roll published two weeks later.That’s nearly a third of the city’s electorate. The Wire has found that the 47.56 lakh names deleted from the list are predominantly working-age men in transient or informal housing, disproportionately migrant workers from Poorvanchal and, on every demographic marker available, blue collar job seekers rather than the settled middle class. This is borne out by testing the data with sex ratio numbers, the age profile, the categories and the surnames.§The Election Commission of India (ECI) has released the Absent, Shifted, Duplicate and Deceased (ASDD) list for all 14,947 booths across 70 assembly constituencies. Each name comes with an EPIC number, a relation type and relation name, the position the person held on the previous roll, their exact age, their sex, and one of four reasons an officer entered against them.To analyse the exercise, The Wire downloaded all 14,947 documents and categorised them. The documents give each person’s exact age, so the 31 to 60 bracket used throughout this piece comes from there. The surname figures come from counting how often each name appears.More men off the listTaking an average across all 70 assembly constituencies, 56.78% of the names removed from the list are men, against 43.21% women. The overall sex ratios of the voter list make this even clearer: Delhi’s roll had 882 women for every 1,000 men when the exercise began. In the SIR, 747 women were deleted for every 1,000 men deleted. In Okhla, 63.68% of the removals were men. In Tughlakabad 61.71%, Moti Nagar 61.51%, and Sangam Vihar, 60.91%.For anyone familiar with Delhi’s urban informal economy, this is perhaps not a surprise. The city’s migrant workforce is overwhelmingly male. Blue-collar job seekers come to the city alone and register at a transient address, a rented room or a bed at a work site, while the family stays in the home state. When the work moves on, a voter registration might be left behind at an address nobody can confirm.Looking at this through the lens of the age profile, between 61% and 75% of removals in any given constituency fall in the 31 to 60 years bracket. Overall, 68.72% of names removed fall within this (working) age range. Deaths account for only 5.94% of the total removals. These numbers suggest it is the working population, not those who died, who are largely being ‘cleaned’ off this list. 91% deletions based on officer not being able to find someoneEvery removal carries one of four reasons, entered by the officer against each name on the booth’s list. Two reasons dominate.The first is ‘Permanently shifted’ accounts for 31,59,223 people, a little over 3.15 million, or 66.42% of the total. The second most common reason is ‘Untraceable or absent’which accounts for another 11,73,552, which is 24.67%. Together, those cover 91.09% of the deletions. Deaths, as we discussed above, account for 5.94%, and duplicate registrations 2.98%.Those with rented accommodation, a large category in a city of migrants like Delhi, are the largest victims of the removals.Interestingly, in the high-volume, migrant-heavy seats, the same three surnames come up every time. In Burari, 10.12% of the removals have the surname ‘Kumar’, 9.56% are ‘Devi’, and 7.54% are ‘Singh’. Narela gives the same three surnames in top deletions, in the same order. In Kirari, 11.82% of the deletions had the surname ‘Devi’, 9.97% were ‘Kumar’, and 5.58% were ‘Singh’.Also read: SIR in Delhi: As One-Third of the Voters Left Out, Field Notes from a Haywire ProcessThese are the most common surnames among Purvanchali workers from Uttar Pradesh and Bihar, who make up much of the tenant base in Delhi’s unauthorised colonies. Across the city, the surnames Kumar, Devi and Singh alone account for 21% of the total removals.So far, this looks like a straightforward story about poor migrant workers living in rented rooms, often temporarily, who have been removed from the voter rolls. Much of the coverage too has framed the matter in this way. Ranking the deletions by the numbers, the worst-hit seats are outer Delhi’s big working-class constituencies. Burari lost 1,67,885 voters, more than anywhere else. Okhla lost 1,63,829, Badarpur 1,46,673 and Vikaspuri 1,46,225.But a constituency’s total number of voters indicates how big it is – and deletions as a proportion of the constituency’s voters paint a different picture. The five biggest seats by booth count lost 1.32 lakh voters on average. The five smallest seats by booth count, lost 42,000. Burari has 409 booths and lost 1,67,885 names. Delhi Cantonment, which lost the fewest voters of any seat at 22,855, has only 94.If we rank the constituencies by average number of voters deleted per booth, this is what it looks like: Delhi averaged 318 removals per booth. Okhla tops the list at 475, Badarpur at 449, Kalkaji and Patparganj at 440. Burari drops from the first to sixth rank. Rohini comes last at 135.Okhla, Badarpur and Tughlakabad are seats which show the maximum numbers of deletions. Both as percentages knocked out and by real numbers. R.K. Puram lost 44.4% of its roll, a larger share than all but two seats in Delhi. Most of the area comprises government staff quarters. New Delhi lost 41.6%, Jangpura 41.4% and Delhi Cantonment 35.4%.Those are among the wealthiest constituencies in India, and all four are dominated by temporary addresses rather than permanent, being government housing, army housing and the South Delhi rental belt. They lost small absolute numbers because they are small seats.Meanwhile Rohini lost 16.5% and Rajouri Garden 20.9%, the two lowest rates in west Delhi. Both are comfortably off and owner-occupied.Some of the poorest parts of the city saw the least percentages of names deleted. Mangolpuri lost 21.2%, Seelampur lost 23.3% and Babarpur 27.0%. These areas are dense, working class, and settled long enough that people own residences they occupy.Tenant versus owner?Delhi is the rare city where this distinction can be tested, because ownership here doesn’t track wealth. The poor often own, through resettlement colonies allotted in the 1970s and ‘80s, plots bought in unauthorised colonies, and ‘lal dora’ village land. The more affluent often rent, because the centre and the south run on government quarters and the rental market. One caveat before the numbers. Nobody publishes ownership rates by assembly constituency in Delhi, so the housing labels used in this piece were assigned seat by seat from what is known about the city’s residential geography. They are judgment, not measurement.The income side can be checked against official estimates – and it has been. MCD assigns every colony in Delhi a property tax category from A to H, with A being the most valuable, and the published list covers 2,638 colonies across 249 wards. Fifty-one of the 70 constituencies share their name with an MCD ward, which allows a comparison.Studying the data in this way, it becomes clear that while both colony value and number of renters can explain the number of voter deletions each seat saw, with number of renters doing the heavy lifting. Constituencies that are renter-heavy saw a higher number of deletions, across income categories.The Wire sorted Delhi’s 70 seats two ways at once, by housing and by income, to see which of the two is actually driving the removals.Sixty-one percent of the poorer constituencies are mostly owner-occupied, against 33% of the richer ones.Even if you look at the data mindful of their respective incomes, renters still remain far more likely to be deleted from the city’s voter rolls. About the data. The ECI published the ASDD list as 14,947 separate booth-wise documents, one per polling booth. Each names the individual electors whose enumeration forms were not received and gives, for every name, an EPIC number, a relation type and relation name, the position held on the previous roll, age, gender, and one of four uncollectable reasons. Figures here come from parsing all of those documents and aggregating them, and the result reconciles exactly with the Chief Electoral Officer’s press note on booth count and all four category totals. Margins come from the February 2025 assembly results.Housing and income categories are not in the published data. There is no ownership percentage behind them and no threshold, because no ownership data was used. Each of the 70 seats was placed in a category by judgment from Delhi’s residential geography, and the coding was not blind to the results. The same finding rebuilt from published figures alone, grouping seats by the male share of their removals and by eastern-plains surname share, gives a spread of 11.4 points against the housing table’s 12.9. Census 2011 ward-level housing tables carry actual ownership percentages and would settle it properly.For our full coverage on the SIR and the issues it has brought to citizens, click here.