Close Contests: 50 Assembly Seats Won by Under 3,500 Votes

Tamil Nadu's 2026 verdict is usually described in round numbers: TVK 108, DMK 59, AIADMK 47. Those numbers are accurate and they are also misleading. They suggest a settled, comfortable result. The margin data tells a different story — a House whose composition rests on a few thousand votes spread across a handful of constituencies, and which could have looked materially different without any change in the statewide mood. Fifty of the 234 seats were decided by fewer than 3,500 votes. Fifteen were decided by fewer than 1,000. Tirupattur, where TVK's R. Seenivasa Sethupathy beat DMK's KR Periakaruppan, was decided by exactly one vote. That is not a rhetorical flourish; it is the recorded margin in the Election Commission's declaration for AC 185. Where the close seats are The tight results are not concentrated in one region or one kind of constituency. They run from Kanniyakumari at the southern tip, where AIADMK's N. Thalavai Sundaram held on by 214 votes, up through Veppanahalli in Krishnagiri, taken by DMK's P.S. Srinivasan by 138, to Polur in Tiruvannamalai, held by TVK by 227. Urban, semi-urban and rural seats appear in roughly the proportion they occupy in the Assembly as a whole. What the close seats do share is a three-cornered structure. In most of them the runner-up was the party that finished second statewide in that region, and a third force — an alliance partner running its own candidate, a strong independent, or a smaller Dravidian party — polled enough to make the top-two contest a coin toss. This is the structural consequence of a genuinely three-bloc election, which 2026 was for the first time since the 1960s. For four decades Tamil Nadu ran a two-front system. Whichever bloc the smaller parties attached themselves to, the contest in almost every constituency came down to a DMK-led front against an AIADMK-led one, and margins reflected that binary: large, stable, and correlated across whole districts. A third bloc with statewide reach changes the arithmetic in a specific way. It does not necessarily reduce the winner's vote; it reduces the vote needed to win. A candidate who would have needed 45 per cent to take a seat in 2021 could take the same seat with 38 in 2026, and the gap between first and second narrows accordingly. Sixty seats were won by fewer than 5,000 votes and 101 by fewer than 10,000. In a state where turnout averaged above 85 per cent and winning candidates routinely polled 80,000 to 100,000 votes, those margins represent leads of two to ten percentage points. By the standards of Indian state elections that is a competitive House, not a comfortable one. Who was closest to whom Of the 50 closest results, TVK won 15 and was the runner-up in 22. That asymmetry matters more than it first appears. It means the marginal seats were not a set of TVK holds waiting to be knocked over; they were mostly seats TVK narrowly failed to take. A uniform swing against every sitting winner — the crudest possible what-if — therefore moves seats towards TVK, not away from it. Run that arithmetic and the direction is consistent. At a 1,000-vote uniform swing, 15 seats change hands and TVK ends up one seat better off, not worse. At 2,000 votes, 27 seats move and TVK gains three. At 3,500 votes — enough to flip every one of the 50 closest seats — TVK gains seven, DMK is unchanged, AIADMK is unchanged, and the losses fall almost entirely on the smaller parties: IUML down two, INC down two, AMMK, CPI(M), VCK down one each. That last row is the finding worth sitting with. The seats most vulnerable to a small swing do not belong to the two large opposition parties. They belong to the small parties whose presence in the Assembly depends on alliance arithmetic in individual constituencies. A shift of two thousand votes per seat, statewide, would not have changed which bloc governs Tamil Nadu. It would have thinned out the Assembly's smaller voices considerably. The 118 line, and why it never moves A majority in a 234-seat Assembly needs 118. TVK holds 108. No swing available in the 50 closest seats takes it past that line on its own — the maximum gain from flipping all 50 is seven seats, which puts it at 115. The government's majority is, and remains, an alliance majority, delivered on the floor of the House by INC, VCK, CPI, CPI(M) and the AIADMK members who crossed over rather than by TVK's own strength. This is why the small-party losses in the swing scenarios matter. The parties most exposed to a modest uniform swing are the same parties whose votes the government needs. A hypothetical Assembly in which IUML, INC, VCK, CPI(M) and AMMK each hold one or two fewer seats is not a marginally different Assembly. It is one in which the floor arithmetic of 13 May 2026 — the 144–22 confidence vote — becomes considerably tighter. What the wide margins say At the other end, the distribution is just as instructive. Edappadi returned AIADMK with 98,110 votes to spare. Sholinganallur and Madavaram returned TVK with margins near 97,000 and 95,000. These are not swing seats under any realistic scenario, and there are enough of them that the overall picture is bimodal: a large block of safe seats for all three major forces, and a thin band of genuinely competitive constituencies that decided which of them formed the government. Safe seats matter for a reason that has nothing to do with the result. They determine how a party's senior figures enter the House. A leader placed in a 90,000-margin constituency campaigns statewide rather than locally; a leader placed in a 3,000-margin constituency spends the campaign at home. The distribution of safe seats across the three blocs in 2026 therefore says something about how each of them expected to fight the election, and about which of them could afford to move its best campaigners around the state. It also sets up the next cycle. A party sitting on a bloc of very wide margins has surplus votes it cannot convert into additional seats, while a party whose wins cluster in the 2,000-to-5,000 band converts votes into seats far more efficiently but has no cushion if the mood moves. In 2026 TVK held both kinds of seat in quantity. That is a strong position, and it is also the position from which a modest adverse swing does the most damage. Half the Assembly — 101 of 234 seats — was won by under 10,000 votes. In a state where average constituency turnout exceeded 85 per cent and winning candidates routinely polled 80,000 to 100,000 votes, a 10,000-vote margin is a lead of roughly five to ten percentage points. By the standards of Indian state elections that is competitive, not comfortable. Reading margins honestly Two cautions apply to any analysis of this kind. First, a uniform swing is a modelling device, not a forecast. Real swings are never uniform; they cluster by region, by urbanisation, by the strength of a particular candidate. A 2,000-vote model swing tells you which seats are structurally fragile, not which seats would actually move. Second, margins describe the past, not the future. Five of these constituencies currently have no sitting member at all, and two more are vacant for unrelated reasons. When by-elections are eventually held, they will be fought on a 2026-27 record rather than a 2026 campaign, with different candidates and, in at least two cases, with the former MLA standing under a different party symbol. There is a third caution, particular to this dataset. A margin is the gap between first and second, and it says nothing about how many candidates were in the field or how the remaining vote was distributed. Two seats with identical 800-vote margins can be entirely different contests: one a straight fight in which the top two took 95 per cent of the vote between them, the other a scramble in which they took 60 per cent and the rest scattered across a dozen names. The first is stable and the second is not, but the margin column treats them as the same. What the margin data does establish, beyond argument, is that the size of the 2026 mandate has been overstated in both directions. It was not a landslide. It was also not an accident. It was a three-way election in which one bloc converted its vote into seats more efficiently than the others, in a set of constituencies narrow enough to fit on a single page.