Misreading the mood: When exit polls get it wrong

When the last vote is cast in an election, public and politicians wait with bated breath for an unavoidable ritual: exit polls.

TV studios and social media whip themselves into a frenzy over the projected figures even as the politicians, especially those predicted to lose, dismiss the numbers.

This is a script that everyone is quite familiar with. Whether we like it or not, exit polls are firmly wedded to the election cycle.

But it is also true that they also get it horribly wrong, with some famous examples being Bihar 2015 and West Bengal 2021 Assembly polls.

A cursory look at around 100 exit polls for 11 Assembly elections since 2015 shows that three in every four were off the mark or nowhere near the actual results. Some of them got the winning party right but the numbers were wrong.

Take the case of 2015 Delhi Assembly elections. None of the seven exit polls that DH analysed predicted a complete sweep by the AAP when it won 67 out of 70 seats. The highest prediction was 53. One of the polls even put the BJP ahead of AAP with 35 seats, when the saffron party won just three seats.

The 2017 Punjab Assembly polls are another example. While one of the three polls predicted the Congress winning around 70 seats, predictions for the AAP were way off the mark, with one giving it as many as 67 seats while two others gave the party around 50 seats. The party eventually won 20 seats.

The 2015 Bihar polls saw just one pollster getting it right as the Grand Alliance trounced the BJP with 178 seats.

In the latest round of exit polls for Assembly polls in five states, most of them got the winners right but when it comes to actual numbers, many of them were off the mark.

While all of them predicted a BJP alliance victory in Uttar Pradesh, only two could get somewhere near the actual figure. In Punjab, nine exit polls were clear about the AAP’s victory but only two could get the extent of the AAP surge.

Why do exit polls go wrong? Is it sampling, inaccuracy in data analysis or external pressure to tweak the results?

“Many of them are conducting opinion polls and showing it as exit polls,” says political analyst and psephologist Bhamidipati Ramamurthy, who previously conducted surveys for the Congress and earlier held positions in the party.

But other pollsters vouched for their methodology.

Ankit Lal, founder of ‘Politique Advisors’ who had previously headed AAP’s social media team, says many times pollsters are not able to capture the last-minute changes in voter behaviour.

Ramamurthy says one needs to study the pattern of constituencies and the state for at least two years before venturing into exit polls. “You cannot just parachute, you need to have the poll rider constantly,” he says.

Animesh Pandey, co-founder of ‘Politique Marquer’ which does exit polls on the brand name P-MARQ, agrees on this point but also stresses on data collection.

"It is important to capture insights from multiple sources and multiple methods. Technology and analytics capabilities of each polling agency also help in fine-tuning the prediction,” he tells DH.

Pollsters put a premium on sampling and the questionnaire. Ramamurthy says even if the methodology adopted is correct, a bad sample can get your exit polls wrong.

Talking about challenges, Abbin Theepura, co-founder of ‘Politique Marquer’, says “huge capital is required to do research, survey and analysis".

The bias of respondents is another issue. "We need to alienate these samples so that the analysis becomes more accurate. Finding the right kind of resources to be deployed on ground, training and monitoring them is a huge task," he says.

It is worrying for pollsters to get numbers wrong but they say in waves like in Punjab, they could only sense it and not predict the actual numbers. Same was the case in Uttarakhand and Goa in 2022, where the BJP won but exit polls showed a close contest.

Pollsters complain that voters do not open up that easily and they sometimes mislead surveyors. Another challenge is capturing the mood of minority voters and women as well as the floating population. Correctly capturing vote share is another concern area for pollsters.

But to blame just the pollster would be erroneous. In one case in the latest exit polls, a political analyst said a channel showed the AAP on the ticker though it is not a player in UP. “This was because a senior editor in the channel has a liking for the party,” he said.

Sometimes, sources said, the numbers are tweaked keeping the advertisement pie in mind. So the channel would not go against the ruling party, fearing that if by chance it returned to power, they could feel the heat.

“Monetary considerations, political affiliations all determine the exit poll results. You need to be dispassionate with data. That is the challenge before a pollster and an editor,” a pollster said.

Watch latest videos by DH here: