From 2021 Bengal To 2024 Lok Sabha, 5 Times Exit Polls Got It Wrong: Why Does This Happen?

Last Updated:April 26, 2026, 20:51 IST

Polling agencies do manage to capture the "mood" of the voter before and after any election in India – they do not, however, always get it right

Predicting the outcome of elections in the world’s largest democracy has long been described as a “psephologist’s nightmare".

But, polling agencies do manage to capture the “mood" of the voter before and after any election in India – they do not, however, always get it right.

There have been instances where the most sophisticated data modelling has failed to anticipate the final result, often rooted in complex demographic shifts and flaws in statistical sampling due to the sheer size and variety of the Indian electorate.

While exit polls are generally decent at sensing the overall public sentiment, they frequently falter when attempting to convert vote percentages into accurate seat counts within India’s “first past the post" system.

Here are five times where exit polls got it spectacularly wrong:

LOK SABHA SHOCKS: 2004 and 2024

2004: The most legendary failure in exit poll history remains the 2004 Lok Sabha elections. Riding on the back of the high-profile “India Shining" campaign, almost every major polling agency predicted a comfortable return for the Atal Bihari Vajpayee-led National Democratic Alliance (NDA).

Many projections suggested the NDA would secure between 240 and 250 seats. In reality, the result was a seismic shift: the NDA was restricted to just 181 seats, while the United Progressive Alliance (UPA), led by the Congress, emerged as the largest bloc, ushering in a decade of government under Manmohan Singh.

2024: Two decades later, the 2024 Lok Sabha elections proved that Indian voters can still baffle the experts. Leading up to the results, most pollsters predicted a landslide victory for the BJP-led NDA with several agencies saying the ambitious “Abki baar, 400 paar" (more than 400 seats) target was achievable.

Predictions generally placed the NDA between 350 and 400 seats. But the reality was far more modest; the NDA stopped at 293 seats, and the BJP lost its individual majority by winning only 240 seats.

The opposition INDIA bloc performed significantly better than expected, securing 234 seats and catching pollsters off guard.

ASSEMBLY MISCALCULATIONS: BIHAR, DELHI, BENGAL

There have been some stark examples of wrong exit poll results in the assembly elections, where local alliances and “waves" can be difficult to quantify.

2015 Bihar: In the 2015 Bihar elections, pollsters struggled to gauge the impact of the “Mahagathbandhan", an alliance between Nitish Kumar’s JD(U) and Lalu Prasad Yadav’s RJD.

Many polls predicted a neck-and-neck fight or a narrow victory for the BJP. Instead, the Mahagathbandhan swept Bihar with 178 out of 243 seats, while the BJP alliance was decimated winning only 58.

2015 Delhi: This was another shocker. Following the 2014 “Modi Wave", experts expected a tight contest between the BJP and the fledgling Aam Aadmi Party (AAP).

While some gave the AAP a slight edge of 35 to 45 seats, none predicted the actual scale of the result: a historic “clean sweep" where it won 67 out of 70 seats, leaving the BJP with just three and the Congress with zero.

2021 West Bengal: This was a high-decibel battle where several pollsters predicted a “dead heat" or even a BJP victory, with some giving the saffron party 120 to 150 seats.

The reality was a massive mandate for Mamata Banerjee’s Trinamool Congress (TMC), which won 213 seats leaving the BJP far below its “200+" target with only 77 seats.

WHY DOES THIS HAPPEN?

One of the primary reasons is the “silent voter" phenomenon. In many regions, particularly those with a history of political intimidation or strong local polarisation, individuals – especially women or marginalised groups – often do not reveal their true intentions to surveyors.

This was identified as a major factor in the failed predictions for both the 2015 Bihar and 2021 West Bengal assembly elections. Sampling bias poses a significant hurdle.

Given India’s massive and diverse population, which is deeply divided by caste, religion, and urban-rural divides, a sample size of 50,000 to 1,00,000 may fail to capture the hyper-local nuances of various communities. There is also the issue of the “undercurrent" – large-scale swings based on local issues that can happen in the final 48 hours of a campaign, which exit polls conducted at the booth may miss.

WHAT ABOUT HISTORICAL ACCURACY?

The biggest failure for polling agencies is the “seat-conversion" gap. In the Indian electoral system, many assembly seats are decided by razor-thin margins, sometimes less than 1 to 2 percent.

Because of this, even a minor 1 percent or 2 percent error in the sample can lead to a massive seat projection error of 30 to 50 seats in larger states. Research into the historical accuracy trends of Indian elections from 2020 to 2024 shows a “hit-or-miss" pattern.

Agencies like Axis My India, C-Voter, and Today’s Chanakya are generally successful at identifying “directional accuracy" – sensing which way the “wave" is blowing or which alliance has momentum. But their performance varies wildly by state.

For instance, accuracy was high in Uttar Pradesh (2022), where most correctly predicted a BJP return, but low in Chhattisgarh (2023), where almost all polls incorrectly predicted a Congress win when the BJP actually secured a clear majority.

SHOULD WE TAKE EXIT POLLS WITH A PINCH OF SALT?

According to experts, the public should treat exit polls as broad indicators of sentiment rather than a definitive scoreboard or final result.

They are considered most reliable when a “poll of polls" – a consolidation of multiple surveys – shows a consistent trend, and least reliable in multi-cornered contests where the vote is split several ways. To maintain electoral integrity, regulatory bodies have stepped in to limit the influence of these polls.

Under Section 126A of the Representation of the People Act, 1951, the Election Commission of India (ECI) strictly prohibits the publication of exit polls during active voting periods. This is intended to prevent the “bandwagon effect", where voters are pushed toward a perceived winner, or the “underdog effect", where sympathy is generated for a perceived loser thereby ensuring that the later phases of an election are not unduly influenced by premature projections.

But the Indian voter will remain an enigma that statistical models are yet to fully master. Until then, exit polls will be fascinating, even if often flawed.

News india From 2021 Bengal To 2024 Lok Sabha, 5 Times Exit Polls Got It Wrong: Why Does This Happen?

Read More