Psephology, in its proper pedigree, is a careful craft. It descends from the calculations of Pierre-Simon Laplace and the scruples of statistics within the sober study of political science. The exit polls India witnessed last week showed psephology as pantomime. Over a dozen organisations including CVoter, Axis My India, Today’s Chanakya, Matrize, CNX and P-MARQ produced seat projections, vote share estimates and detailed caste-wise breakdowns across states as diverse and complex as West Bengal and Tamil Nadu.
The running theme was near-identical. The BJP was positioned for substantial gains across multiple states. Some agencies appeared a full day after the final phase of polling concluded not one of these agencies disclosed its ownership structure, named its institutional backer, declared its client relationships with state governments or political parties, or published its methodology. They appeared, pronounced and disappeared to be recalled by the same channels at the next election, regardless of accuracy.
Their chief qualification seemed less methodological mastery and more media malleability: the readiness to rehearse results before the real results arrived. The historical record on accuracy is not flattering. In the 2021 West Bengal elections, the aggregated poll of polls estimated the TMC would win approximately 156 seats and the BJP approximately 121 in a 294-seat assembly. The actual result was TMC 215, BJP 77. The exit polls underestimated the TMC by about 60 seats and overestimated the BJP by more than 40—a collective structural failure, not a marginal rounding error. In the 2024 Haryana elections, every major agency projected a decisive Congress majority of 44-65 seats with the BJP trailing at 15-28. The BJP won 48 seats and returned to power for an unprecedented third consecutive term.
In the 2024 Lok Sabha elections, multiple agencies predicted the NDA would cross 350 seats. The BJP ended with 240. The consistency of these failures reveals a pattern more troubling than random inaccuracy. India’s exit polls demonstrate a documented tendency to cluster projections in a manner the British polling inquiry identified as herding, whereby pollsters consciously or unconsciously adjust their numbers towards each other to avoid the professional embarrassment of standing alone with a contrarian prediction.
Several agencies this week released precise caste and community voting patterns for Bengal and Tamil Nadu, with exact percentages of Scheduled Caste voters, OBC preferences and minority behaviour. This data does not officially exist. No credible institution maintains a verified real-time caste database accessible in the hours immediately following polling. When an exit poll agency publishes such figures, it is fabricating a sociological narrative in numerical disguise and broadcasting it to an audience with no mechanism for verification.
In Britain, organisations that publish political polls must be members of the British Polling Council, which mandates full disclosure of methodology, sample size, weighting procedures, fieldwork dates and commissioning identity. Full data tables must be published within two working days. When British pollsters failed in the 2015 general election, the British Polling Council and Market Research Society commissioned an independent LSE-led inquiry that produced 12 methodological reforms which the industry adopted. In the US, the National Council on Published Polls requires disclosure of sample sizes, margins of error, question wording and funding sources, with professional consequences for non-compliance.
In India, no such principle operates. Beyond the Election Commission’s prohibition on publishing exit polls until voting concludes, there is no registration requirement, obligation to publish methodology or a professional body with enforceable standards. The same agencies that missed Bengal 2021 by 60 seats were invited back for Lok Sabha 2024. The same agencies that declared Congress victorious in Haryana were projecting outcomes for Tamil Nadu.
Failure is not merely tolerated. It is rewarded with renewed commissions and expanded airtime, because TV channels are not purchasing accuracy. They are purchasing content that fills the vacuum between the conclusion of voting and the declaration of results while inflation, the plummeting rupee and the energy crisis wait patiently outside the studio.
Tomorrow, when ballots are counted across Bengal, Tamil Nadu, Kerala, Assam and Puducherry, the reckoning arrives conducted by the most impartial examiner available to a democracy, which is reality itself. Examine this week’s projections carefully. Every agency has shielded its predictions behind arithmetically improbable ranges, projecting not a number but a spread wide enough to render the prediction technically unfalsifiable. An agency projecting the BJP at between 130 and 175 seats in a 294-seat assembly cannot be wrong if the party wins 131 or 174. An agency projecting the TMC at between 100 and 155 has covered a range capacious enough to contain almost any conceivable result. These ranges are escape hatches engineered to ensure that no matter what counting reveals on Monday, the agency can point to some corner of its projection and declare vindication.
But Monday’s counting will expose not merely whether an agency correctly identified the winning party, but whether its seat estimates bore any relationship to the actual distribution of victories across constituencies, and whether its fabricated caste-wise data had any empirical correspondence to what actually happened at the ballot box.
There is, moreover, a more sinister question that Monday’s results may illuminate. When a dozen agencies independently arrive at virtually identical seat projections for five elections covering crores of voters, one of three explanations must hold. Either rigorous independent survey produced identical conclusions by coincidence statistically implausible. Or several agencies worked from shared data of undisclosed provenance, raising the question of whose data it is. Or some projections were designed to reflect a desired outcome rather than a reality measured with psephology.
May 4 will not definitively resolve which explanation holds. But it will provide the evidentiary foundation for a more honest public reckoning with the question than India has so far managed. Exit polls that shroud their ownership, conjure their caste calculus and engineer elastic ranges pollute democratic discourse. India’s elections—vast, volatile and vulnerable—are among the most intricate exercises in modern democracy. They unfold not in studios but in sun-scorched streets and rain-lashed rural routes, where voters queue with a patience that puts punditry to shame.
Across constituencies that cameras seldom capture and commentators rarely comprehend, choices are made with quiet conviction. To process that collective will through a commercial contraption is to trade democratic dignity for televisual drama. It is, at its irreducible essence, contempt for the voter dressed in the borrowed vocabulary of science. Monday’s counting will not cure that contempt. But it may finally make it impossible to deny.
Read all columns by Prabhu Chawla
Prabhu Chawla
prabhuchawla@newindianexpress.com
Follow him on X @PrabhuChawla