Opinion polls and exit polls are Indian elections’ favourite sideshow — and their most controversial. Pre-election surveys predict winners; exit polls, conducted as voters leave booths, claim near-certainty. Sometimes they are spectacularly right, sometimes spectacularly wrong, and the Election Commission bans their publication during multi-phase voting. This explainer covers how polls work, why they err, and what the bans are for.

How are opinion polls conducted?

Serious polling is applied statistics. Agencies draw stratified random samples — a few thousand to tens of thousands of respondents across regions, castes, genders and urban-rural divides — and weight results to match the electorate’s demographics. Face-to-face interviews remain the gold standard in India, where phone penetration skews samples. The best pollsters (CSDS-Lokniti’s academic surveys) publish methodology; many media polls do not. Converting vote shares to seats adds another layer of modelling, since India’s first-past-the-post system makes seat predictions far harder than vote estimates.

Why do polls get it wrong?

Indian polling has famous misses: 2004 (the BJP’s India Shining defeat), 2015 Bihar, 2021 Bengal, and the 2024 Lok Sabha, where most exit polls wildly overestimated the BJP. The causes are structural. Sampling: India’s diversity defeats small samples; marginalised groups are under-reached. Shy voters: respondents hide preferences, especially for stigmatised choices. Late swings: Indian voters decide late, after polling ends. Herding: agencies tweak results toward consensus to avoid outlier embarrassment. Sponsorship bias: media-commissioned polls face pressure for dramatic narratives. And the vote-to-seat translation multiplies every error.

What is the difference between opinion and exit polls?

Opinion polls survey voting intentions before polling day; exit polls interview voters after they vote, asking whom they chose. Exit polls should be more accurate — the act is done — but Indian exit polls have their own pathologies: interviewers cluster at accessible booths, voters misreport, and the record is checkered enough that 2024’s debacle triggered industry-wide soul-searching. Post-poll surveys (conducted days after voting, like CSDS’s) are methodologically strongest but arrive too late for television drama.

Why does the EC ban poll publication?

Section 126A of the Representation of the People Act bans exit-poll publication from the first hour of polling until half an hour after the final phase ends — in multi-phase Lok Sabha elections, a weeks-long blackout. The rationale: early results could influence later-phase voters, violating the level playing field. Opinion polls face no such ban, though demands for one recur after every miss. Critics of bans cite free speech; defenders cite voter manipulation. The debate replays each election cycle, unresolved.

• Reading polls wisely: check sample size, methodology, sponsor and track record — not just the headline numbers.
• Margins matter: a 3% vote-share error can flip dozens of seats under first-past-the-post.
• Trends over points: direction of movement is more reliable than precise seat counts.
• 2024’s lesson: even unanimous exit polls can be wrong; treat election-night graphics as entertainment with statistics.

Polls are imperfect instruments measuring a moving target — India’s billion-strong electorate in all its complexity. Used humbly, they illuminate; hyped as prophecy, they mislead. The wise viewer watches the methodology slide, not just the winner’s graphic.

The business of polling

Election polling is also big business with perverse incentives. Television channels commission polls as programming — the graphics, the countdowns, the debates drive ratings for weeks. Agencies compete for contracts, and the temptation to deliver sponsor-friendly numbers is real; several sting operations have caught pollsters offering to tailor findings. The lack of a self-regulatory code — no published standards, no accreditation — lets fly-by-night operators flourish alongside serious researchers. Until the industry cleans itself up, viewers should treat every poll as advertising until proven otherwise.

Polls and the bandwagon effect

Do published polls change votes? The bandwagon theory says voters flock to projected winners; the underdog theory says they sympathise with projected losers. Indian evidence is thin but suggestive: the EC’s ban assumes influence exists, and parties certainly behave as if it does — leaking favourable internal polls to build momentum. Whether polls merely measure or actively manufacture outcomes remains unproven, but their power to set narratives — who is winning, who is finished — is undeniable, making their accuracy a democratic concern, not just a statistical one.

FAQs

What is the difference between opinion and exit polls?

Opinion polls measure voting intentions before election day; exit polls interview voters after they cast their ballots, asking whom they voted for.

Why are exit polls banned during elections?

Section 126A bans publication during multi-phase polling so early results cannot influence voters in later phases.

Why did 2024 exit polls fail?

Sampling gaps, shy voters, late swings and vote-to-seat modelling errors combined to overestimate the BJP’s tally dramatically.

Source: CSDS-Lokniti