In a first-past-the-post electoral system, the composition of the voter roll is not incidental to democracy—it is foundational. Tamil Nadu’s 2026 assembly election results offered a stark reminder: one constituency was settled by a single vote. In Maharashtra, the Mumbai Northwest Constituency seat was decided by a thin margin of 48 votes in the 2024 parliamentary election. The principle is elementary but consequential: each name on the roll, and each name removed from it, carries potential electoral weight.

This article’s purpose is the important one of establishing what the roll revision—taken on its own terms—was arithmetically capable of doing.

A prefatory note on the scope of this analysis is necessary. The exercise is strictly arithmetical in character: it asks whether the scale of deletions and additions from electoral rolls under the Special Intensive Revision (SIR) was sufficient, in quantitative terms, to have affected seat outcomes in the 2026 West Bengal Assembly elections.

It does not model political variables such as shifts in voter preference, the performance of the state government, candidate effects, or the role of community mobilisation. These factors are consequential and have been addressed elsewhere. This article makes no claim to offer a comprehensive explanation of the 2026 electoral outcome. Its purpose is the narrower but important one of establishing what the roll revision—taken on its own terms—was arithmetically capable of doing.

It is against this backdrop that West Bengal’s SIR demands careful examination. Across the state, 90.82 lakh voter names were removed from the electoral roll. Of these, 58.18 lakh were struck off under the “Absent, Shifted, Dead, and Duplicate” category during the draft revision; a further 27.16 lakh were excluded following judicial review of “Under Adjudication” cases, and 5.46 lakh as “Draft Roll Deletions”.

Against these deletions, 1.88 lakh new names were incorporated into the final roll published on 28 February 2026. Yet, by mid-April—just days before Phase I of polling—the figure had risen to 6.96 lakh, implying a surge of nearly 5 lakh names within a single month. In aggregate, approximately 7 lakh new voters were enrolled ahead of the assembly elections.

Crucially, when 5 lakh names were added on 18 April 2026, the Election Commission of India (ECI) did not release a constituency-level breakdown, in contrast to the detailed data provided for 90.82 lakh deletions and 1.88 lakh additions. The Supreme Court on 20 April 2026 declined to hear an oral petition challenging these Form 6 additions made after the SIR was formally concluded.

This is not an assertion that outcomes were determined solely by the SIR. It is an observation that roll revision entered the zone of electoral competitiveness in a manner that warrants rigorous scrutiny.

Neither a demographic breakdown by age or gender nor a count of first-time electors among the new additions has been placed in the public domain. ECI officials state that while the total figures meet standard rules, detailed data is stored separately and may be released later.

During the SIR proceedings before the Supreme Court, Justice Joymalya Bagchi offered a pointed observation from the bench: “If 10% of the electorate does not vote and the winning margin is more than 10% …what will happen? Suppose the margin is 2% and 15% of the electorate who are mapped could not vote, then maybe …we are not expressing any opinion, but we would definitely have to apply our minds.” Constituency-level data subsequently released suggest these concerns were well-founded.

Constituency-level Data

What the constituency-level data reveal is that the SIR was far from a politically neutral administrative exercise. Its significance rests on a single structural fact: across a large portion of West Bengal’s assembly, the volume of voter deletions was sufficient, arithmetically, to alter the result.

This is not an assertion that outcomes were determined solely by the SIR. It is an observation that roll revision entered the zone of electoral competitiveness in a manner that warrants rigorous scrutiny. The concern is sharpest in the Under Adjudication category: constituency-level analysis by the SABAR Institute found that in Nandigram, 95.5% of voters deleted under Under Adjudication were Muslim, despite Muslims comprising only 33.3% of the Absent, Shifted, Dead, and Duplicate list and approximately 23% of the constituency’s total population.

A corresponding analysis for Bhabanipur showed that 40.1% of Under Adjudication deletions were Muslim voters, against a Muslim share of 22.7% in the Absent, Shifted, Dead, and Duplicate list and roughly 20% of the population.

More Deletions than Margins

The most direct measure of the SIR’s electoral footprint is the number of seats in which gross deletions—combining Absent, Shifted, Dead, and Duplicate, Under Adjudication, and Draft Roll Deletion removals—exceeded the final winning margin. Aggregating these three deletion streams, the total number of removed voters surpassed the winner’s margin in 160 constituencies. West Bengal has 294 seats in total, which means these constituencies account for 54.42% of the assembly.

Figure 1: Deleted Voters More Than the Winning Margin in 160 Seats

Of these 160 constituencies, the Bharatiya Janata Party (BJP) won 105 and the Trinamool Congress (TMC) 53, with others accounting for two. The contrast with 2021 is dramatic: in these same seats, the TMC had won 140 and the BJP 20.

Three districts bore a disproportionate share of the roll contraction, producing the largest clusters of seats where gross deletions exceeded winning margins: South 24 Parganas (19 of 31 seats), North 24 Parganas (27 of 33 seats), and Murshidabad (16 of 22 seats). In North 24 Parganas, the TMC had been dominant in 2021, winning 24 of the 27 affected seats; in 2026, the BJP overturned this, capturing 21 of them. In South 24 Parganas, the TMC had swept all 19 such seats in 2021, but the BJP won 10 of them following the SIR.

In Jangipara, a BJP margin of 862 coexisted with 17,541 total deletions, of which 5,432 were Under Adjudication.

Beyond these epicentres, the roll contraction cut deeply into Muslim-majority and competitive districts. In Murshidabad, the TMC’s 2021 count of 14 from 16 affected seats collapsed to 7, with the BJP and Congress each picking up 7 and 2, respectively. In Purba Bardhaman, 11 of the TMC’s 13 previously held affected seats shifted to the BJP. Across Howrah and Hooghly combined—25 affected seats that the TMC had won in 2021—the BJP claimed 15 in 2026.

Urban constituencies were equally affected. In Paschim Bardhaman, the BJP won all nine affected seats, taking six that had been TMC territory in 2021. In the combined Kolkata North and South districts, the BJP captured six of 11 affected seats—including Bhabanipur, where Chief Minister Mamata Banerjee fell to Suvendu Adhikari. The geographic pattern is unambiguous: the roll contractions fell most heavily on constituencies the TMC had held most securely in 2021, allowing the BJP to convert 105 of these 160 deletion-heavy seats.

Individual constituencies render this abstraction concrete. In Satgachhia, the BJP’s winning margin was 401 votes, against a total of 27,087 deletions—comprising 17,671 Absent, Shifted, Dead, and Duplicate, 8,785 Under Adjudication, and 631 Draft Roll Deletions. In Rajarhat New Town, a BJP margin of 316 stood beside 64,980 total deletions, of which 24,132 fell under Adjudication. In Raina, the BJP’s 834-vote margin accompanied 23,312 total deletions, including 11,284 under Adjudication. In Jangipara, a BJP margin of 862 coexisted with 17,541 total deletions, of which 5,432 were Under Adjudication.

Absent, Shifted, Dead, and Duplicate Deletions

Isolating Absent, Shifted, Dead, and Duplicate deletions and setting Under Adjudication aside entirely, the political asymmetry persists. In 109 constituencies, routine Absent, Shifted, Dead, and Duplicate removals exceeded the victory margin. The BJP won 71 of these seats, nearly double the TMC’s 36, with the Congress taking two. Geographically, the pattern is consistent: North 24 Parganas (18 seats) and South 24 Parganas (16 seats) lead the list of affected areas.

Figure 2: Deleted Absent, Shifted, Dead, and Duplicate Voters More Than Winning Margin in 109 Seats

In 2021, the TMC won 101 of them compared to just eight for the BJP. To isolate the direct change, the data show that 66 of these 109 seats actually changed hands between the two elections. The TMC lost all 66 flipped constituencies, with the BJP capturing 64 and the Congress gaining the remaining two. Ultimately, the data demonstrates that routine Absent, Shifted, Dead, and Duplicate deletions were undeniably concentrated in seats the TMC had previously won, and in areas where these roll clean-ups outnumbered the winning margin, the TMC’s dominance was effectively shattered.

Taken together, the twin districts of North and South 24 Parganas accounted for nearly a third of all seats heavily impacted by the Absent, Shifted, Dead, and Duplicate deletions. The TMC had nearly swept North 24 Parganas in 2021, winning 17 of 18 seats. But the situation flipped in 2026 with the BJP capturing14 of them, reducing the TMC to just four. In South 24 Parganas, the TMC’s unblemished 16-0 record since 2021 was dismantled as the BJP captured eight of them following the revision.

In the combined Howrah and Hooghly, the TMC had held all 20 Absent, Shifted, Dead, and Duplicate-affected seats in 2021. In 2026, the BJP won 13. In Murshidabad, where the Muslim population is substantial, the TMC’s 2021 tally of nine out of 10 such seats fell to just two—the BJP took six and the Congress two.

Under Adjudication Deletions

The Under Adjudication category exhibits a qualitatively distinct and more politically sensitive pattern. In 49 constituencies, Under Adjudication, deletions alone exceeded the winning margin. The shift relative to 2021 is striking: the TMC had held 48 of these 49 seats; the BJP held only one. In 2026, the distribution inverted—the BJP won 26, the TMC 21, and Congress two. A key demographic feature of these 49 seats is that the average Muslim population share stands at 38.34%, well above the state average of 27%.

Figure 3: Deleted Under Adjudication Voters More Than Winning Margin in 49 Seats

Murshidabad was the most affected district, with eight seats in which the Under Adjudication exceeded the winning margin. Other affected districts included North 24 Parganas, South 24 Parganas, Purba Bardhaman, and Hooghly.

The overall picture is one of compound vulnerability. Deletions of Under Adjudication entries, concentrated in constituencies with high Muslim populations and historically dominated by the TMC, intersected with razor-thin winning margins to produce outcomes in which the TMC lost more than half its strongholds in this category to a party that had barely registered in these seats five years earlier.

The constituency-level figures are instructive. In Samserganj, the TMC prevailed by 7,587 votes, yet the under-adjudication deletions totalled 74,775—nearly 10 times the margin. In Raninagar, the Congress won by 2,701 votes, while 17,140 Under Adjudication voters were removed. In Jangipur, with more than 57% Muslims, the BJP’s margin was 10,542 against under-adjudication deletions of 36,581. In Raina, the BJP’s 834-vote margin sat alongside 11,284 Under Adjudication deletions, a disparity that illustrates how far administrative action outpaced the electoral signal.

The individual identities of voters deleted under the Under Adjudication category have not been publicly disclosed. However, constituency-level ECI data reveal a striking structural pattern.

Figures 4 and 5 are two scatter plots drawn from ECI data. Figure 4 shows that Absent, Shifted, Dead, and Duplicate deletions—carried out in December 2025 via physical mapping—decline as the share of the Muslim population at the constituency level rises. This is broadly consistent with patterns of demographic mobility and natural attrition.

Figure 5, however, shows the reverse: Under Adjudication, deletions increase as the Muslim population share rises. The Under Adjudication process, conducted as a supplementary deletion within the SIR framework, therefore moved in the opposite demographic direction to the standard Absent, Shifted, Dead, and Duplicate exercise.

No administrative rationale has been offered for this divergence. As noted, the average Muslim population share across constituencies where Under Adjudication-deleted voters exceed the winning margin stands at 38.4%. Micro-level investigations by the SABAR Institute found that in Nandigram, 95.5% of Under Adjudication-deleted voters were Muslim despite the community comprising about 23% of the constituency’s population; in Bhabanipur, the figure was 40.1% against a Muslim share in the population of roughly 20%.

Figure 4: Relation Between Absent, Shifted, Dead, and Duplicate Deletion and Constituency-wise Muslim Population

Figure 5: Relation Between Under Adjudication Deletion and Constituency-wise Muslim Population

None of this is to argue that political change in West Bengal since 2021 is illusory, or that every deleted voter would have voted against the eventual winner. But when deletions occur at a scale of 90.82 lakh, when Under Adjudication deletions are concentrated in Muslim-majority constituencies, and when results are routinely decided by margins of a few hundred or a few thousand votes, the question of how many outcomes would have differed without the Under Adjudication exercise cannot be avoided.

Unresolved Data Gap

Between the finalisation of the SIR roll and the close of nominations, the ECI enrolled approximately 6.96 lakh new voters in West Bengal, including around 5 lakh Form-6 applicants. Roughly 3.22 lakh of these were eligible in Phase I (23 April 2026) and 3.88 lakh in Phase II (29 April 2026).

At the district level, North 24 Parganas registered the highest additions at over 71,000, followed by Kolkata at nearly 44,000; Howrah, Hooghly, and East Midnapore also saw significant increases.

An assembly-constituency-level breakdown of 1.88 lakh newly added voters on 28 February 2026 was available. But strangely, no assembly-constituency-level breakdown of 5 lakh newly added voters in April 2026 has been released— neither by the Chief Electoral Officer (CEO), West Bengal, nor by any independent analysis. No age, gender, or first-time-voter classification has been made available.

Petitions seeking constituency-level disclosure were declined by the Supreme Court, and the ECI has taken no further steps. Given that the statewide vote-share gap between the BJP and the TMC is approximately 32 lakh votes, the 5-lakh figure is non-trivial. The continued opacity around these additions constitutes a significant gap in the public record of this election.

Simulating a No-SIR Election

To assess the SIR’s electoral weight with greater precision, I construct two hypothetical no-SIR scenarios. Neither purports to reconstruct individual voter behaviour with certainty. Each offers a structured, possible mathematical impact test of how sensitive seat outcomes were to the scale of deletions.

The analysis computes the total SIR impact—combining Absent, Shifted, Dead, and Duplicate, Under Adjudication, and Draft Roll Deletion deletions and 1.88 lakh inclusions in February 2026—and compares it to the actual winning margin in each constituency. Because the final inclusion of 5 lakh voters (in April 2026) at the constituency level was not released by the ECI, we could not incorporate that figure into our simulation model.

Counterfactual Case 1

The first scenario adopts the most conservative assumption for the winning party in 2026: every deleted voter, had they been retained, would have voted for the runner-up; and every newly added voter among the 1.88 lakh included, had they not been included, would have voted for the winner.

Of the 136 seats that changed hands between 2021 and 2026, the BJP captured 131. Within these 131, the data identify 88 seats where the winning margin was smaller than gross deletions, and four seats where the winning margin was smaller than the new additions from the 1.88 lakh included. In all 92 of these cases, the TMC had held the seat in 2021; in 2026, the BJP won 90 and the Congress won two.

Had the SIR not occurred, and accepting the worst-case assumption about deleted and included voter preferences, the TMC could theoretically have defended all 92 of these seats.

Under this worst-case counterfactual, all 92 seats would theoretically have remained with the TMC. In reality, 90 were won by the BJP in 2026 and two by the Congress—a near-complete reversal of the 2021 picture in this subset of constituencies.

Had the SIR not occurred, and accepting the worst-case assumption about deleted and included voter preferences, the TMC could theoretically have defended all 92 of these seats. Added to its actual tally of 80 seats, that would have yielded 172 seats—enough to form the government. The BJP, losing 90 of these seats, would have been reduced to approximately 117.

Figure 6: Electoral Outcome with SIR and without SIR (Case 1)

SIR Impact

To illustrate the depth of roll modification at the individual constituency level, I define “SIR impact” as the excess of gross deletions over the winning margin. In 63 constituencies, the SIR impact exceeded 10,000. In 39, it exceeded 20,000. In 21, it exceeded 30,000. Table 1 presents five constituencies where the SIR impact reached 45,000 or more, each a seat where the TMC was unseated by the BJP.

Table 1: Five Constituencies where SIR Impact was more than 45,000 and where BJP replaced TMC

Case 1 is, by construction, a sensitivity measure rather than a prediction of actual voter behaviour. Yet the numbers point to a clear reality: roll revisions of this magnitude were not administrative housekeeping. In constituency after constituency, the volume of administrative churn vastly exceeded the electoral signal it surrounded.

Counterfactual Case 2

The second scenario is methodologically more cautious. Rather than assuming that all deleted voters would have opposed the winner and all included voters would have opposed the runner-up, it redistributes them in proportion to each constituency’s party-wise vote shares—using 2021 vote shares for deletions and 2026 vote shares for additions. In the case of deletions, the 2021 election is treated as the best available estimate of how those voters might have behaved; in the case of additions, the 2026 election is treated as the best available estimate.

The simulation proceeds in three steps.

First, all 90.82 lakh deleted voters—comprising 58.18 lakh Absent, Shifted, Dead or Duplicate, 27.16 lakh Under Adjudication, and 5.46 lakh Draft Roll Deletions—are reinstated into the constituency-level electorate, scaled to the percentage of votes cast in that constituency in the 2026 election.

Second, these reinstated voters are allocated to parties in proportion to each constituency’s 2021 vote-share distribution, on the assumption that their preferences mirror those of the prior electorate.

Third, the 1.88 lakh additions are removed using the 2026 party distribution, to further sharpen the counterfactual. It would have been preferable to remove the entire 7 lakh addition, had the ECI published constituency-level data on the remaining 5 lakh additions as well. As noted, no such breakdown was released for those additional 5 lakh voters, despite multiple petitions.

Figure 7: Electoral Outcome with SIR and without SIR (Case 2)

Under this simulation, the BJP would have won 166 seats, the TMC 121, and others six—across the 294 valid constituencies. The BJP retains a clear majority and its mandate to form the government. The overall verdict is thus preserved. However, the scale of the BJP’s majority is substantially reduced: from the actual 207 seats to a no-SIR tally of 166.

This scenario does not claim to reconstruct actual voter behaviour; it is one plausible no-SIR outcome. The arithmetic suggests that the deletions, and arguably the additions too, were sufficiently large and politically concentrated to plausibly shift outcomes across a significant swath of the assembly.

Specifically, 77 constituencies would have recorded different results: 59 flipping from the BJP to the TMC, and 18 from the TMC to the BJP—a net transfer of 41 seats away from the BJP, yielding BJP 166, TMC 121, and others six. However, this tally does not account for the impact of the inclusion of 5 lakh voters at a very late stage in April 2026, for which the ECI did not provide any constituency-wise breakup. If we had the breakup of these 5 lakh voters, it could only have further reduced the BJP’s winning tally.

SIR as an Electoral Actor

Two no-SIR scenarios yield distinct outcomes. Under Case 1, which assumes all deleted voters would have supported the runner-up and all newly added voters would have supported the winner, the TMC would have won approximately 172 seats and formed the government; the BJP would have been left with around 117.

Under Case 2, which redistributes deleted voters according to 2021 vote shares and redistributes only part (1.88 lakh) of the newly added 7 lakh voters according to 2026 vote shares, the BJP would still have formed the government with 166 seats, against 121 for the TMC and six for others. The first scenario overturns the result; the second would possibly preserve it but diminishes its scale.

What this analysis demonstrates is that the SIR—operating at the scale it did—was arithmetically capable of amplifying whatever anti-incumbency existed, translating what might have been a political shift into a legislative supermajority.

What the data collectively demonstrate is threefold. The Absent, Shifted, Dead, and Duplicate deletions were concentrated in seats the TMC had held most securely, and wherever those deletions exceeded the margin, the TMC’s incumbency was undermined. Under Adjudication, deletions followed a community-specific trajectory—rising with Muslim population share—with the sharpest impact in minority-heavy constituencies.

The opacity surrounding 7 lakh post-SIR additions leaves an unresolved gap in the electoral record. In a state where victory margins are frequently narrow, an SIR exercise of this magnitude—90.82 lakh deletions and roughly 7 lakh additions—was not a procedural footnote to the election. Against a statewide BJP-TMC vote gap of approximately 32 lakh, the roll revision was, by its own statistical weight, a political actor in its own right.

A final qualification deserves emphasis. This analysis does not claim that anti-incumbency sentiment and popular discontent with the TMC government were absent or irrelevant to the 2026 result. On the contrary, there is considerable evidence that governance fatigue, perceptions of administrative overreach, and the erosion of the TMC’s social coalition had independently weakened the party’s electoral position—forces that no roll revision exercise created.

What the present analysis demonstrates, however, is that the SIR—operating at the scale it did—was arithmetically capable of amplifying whatever anti-incumbency existed, translating what might have been a political shift into a legislative supermajority.

While the SIR did not create discontent out of nothing, it heavily amplified its impact. By removing 90.82 lakh voters primarily from TMC strongholds and Muslim-majority pockets, the revision vastly magnified the electoral consequences beyond what the underlying political tide alone would have achieved.

Subhasish Dey (subhasish.dey@warwick.ac.uk) is an Associate Professor of Economics at the University of Warwick.