The first past the post model is often the nightmare of a psephologist in India. It leads to vagaries of the kind the normal mind can least imagine. Chitrasen Sinku won the Singhbhum Lok Sabha seat in Jharkhand in 1996, polling just 21 per cent votes. The example of this Lok Sabha seat when scaled up to an entire state exemplifies the predicament over Vijay's vote share conversion in Tamil Nadu.
A Wholesome Psephological Framework For The First Past The Post Model
Can a vote share of even just 25 per cent, not the peak of 35 per cent as estimated by some polls for the TVK convert to a big win?
A framework of three factors explains this.
1. Concentration of votes: Breaks up the states into different categories of seats, which are not necessarily contiguous, but that which correspond to similar patterns of contest. In the case of Vijay, these, as per ground feedback, could be portions of northern Tamil Nadu, urban areas, minority-dominant and Scheduled Castes (SC) seats. Think of it like this - if this set of non-contiguous seats forms a smaller state by itself, say with 150-160 seats, the votes here could be concentrated enough to deliver the state to a lucky winner. The estimation of this catchment landscape is the first key.
2. Effective number of political parties on the landscape: Simply put, this means 'what is the real polarity of the contest'? Are they two parties, 2.5 parties, triangular, 3.5 parties, etc? For each concentration region on the landscape, this has to be estimated before you use the first past the post model to make forecasts.
3. Effective modeling of multi-party swings: Almost with an engineering precision you need to understand how the multiple flows of votes between different parties contribute to the new vote on each party individually.
Elections with multiple polarities and sudden surge or growth of one particular party have often caught psephologists on the wrong foot because one of the above steps goes amiss. They simply follow the law of large numbers.
What Is The Application Of The Law Of Large Numbers?
It simply assumes that if we poll a large number of respondents in a survey, they capture the overall trend and those will then be used to project a seat-by-seat result.
Typical cases of multipolarity and sudden surge that have been largely wrongly estimated in the past:
1. Tamil Nadu 1998
2. Maharashtra, Andhra Pradesh and Tamil Nadu 2009
3. Delhi 2013 and 2015
4. UP 2014, 2017 and 2024
5. Haryana 2024
Most of the pollsters got the above wrong as they simply followed the law of large numbers.
The Case Of Tamil Nadu Assembly Polls 2026
Why Tamil Nadu might have been a nightmare for many a psephologist is that in multi-cornered contests and a new party in place, estimates made without the multi-party flow creation and simply following the law of large numbers is inadequate. You may not get it wrong all the time, but surely you won't be right most of the time, and won't have the confidence to say where you went wrong.
So let us try and understand what is creating a wide range of results:
1. Inaccurate segmentation of the state into categorised seats and regions based on polarity of the contest. You would fail to convert vote shares accurately into seat shares if you do not have this. This would create gross level errors, as the overall vote shares assume a homogeneity that does not exist on the ground. This creates wide variations such as some polls suggesting the party. A winning and another poll suggesting a diametrically opposite party B winning.
2. Inaccurate multi-party flows: You may end up getting the direction of the result correct e.g. surge in TVK voters, but you won't get the order of the vote shares right, so you can't exactly call the winner. You may have polls which are directionally aligned, but the value of vote share and its conversion to seats would vary from poll to poll. This has been observed in the case of two polls giving a substantial vote share to the TVK, but being inconsistent in terms of the seats they gave to the TVK.
In the context of Tamil Nadu, the following:
Situations Where Accurate Segmentation Of The State Matters Of Vijay As Splitters Of Vote
Given that Vijay has a subtle appeal cutting across religious lines, the DMK has tried to ensure that it becomes the singular choice of the minority voters. The party has empaneled support of various Christian denominations, with apparent circulated letters from the church, which state that the DMK has done enough for minorities not to waste their vote, including in seats like Perambur from where Vijay is contesting. Similarly, SCs are spread across the state in a variety of clusters. Non-contiguous clusters and regions based on these would help in working on the varying polarity of the contest, across the state.
Modelling The Effective Polarity Of The Contest
With the advent of Vijay there has been a race to the bottom to salvage every single seat by both the alliances, some crafted just for a couple of seats, and some crafted for a couple of districts. While these look cosmetic, they actually alter the winners threshold often in multi-party scenarios.
The DMK for example has in its alliance four Muslim-centric parties: MMK, MJK, SDPI, and Muslim League, often with leaders with a hold over just a couple of seats, but an engagement that will ensure minimal splintering of votes, helping it beat the threshold.
Similarly, the DMDK, a party with just 0.42 per cent votes in the last assembly polls, was given a generous 10 seats by the DMK alliance, which is a tacit acknowledgement that the party was trying to go past the rubicon even if in just two or three districts by pooling in from another party.
Modelling Multi-Party Flows
The hypothesised threat to the DMK in urban areas from Vijay could be real, especially in areas where the minority vote along with the urban youth vote could upset the traditional apple cart of the DMK as well as the AIADMK. If the skew in Vijay's voters is more urban, and the VCK is taking away more votes from the AIADMK as well as DMK, it could render the AIADMK third.
Only Monday would tell how effectively each of the psephological agencies used this edifice. I expect most experiences to be sobering.
(The author is co-founder of VMR [Votersmood Research] and founder of M76 Analytics. You can follow him @JMAI799249 on X)
Disclaimer: These are the personal opinions of the author
