Lok Sabha Elections: How do we predict the exit poll numbers?

how do you read these exit polls which released Saturday evening? (Representative image)

The exit poll aired on various channel last evening suggests a general outcome, the BJP seems to be crossing the 300 mark and National Democratic Alliance (NDA) is certainly scoring 350 seats, though some exit polls have also estimated 400+ figures for the NDA.

While the estimates of the exit poll may not be on the expected lines for many, but reactions to these exit polls were on expected lines—some say--they do not believe these numbers. Some (strong critics of BJP) alleged that these exit polls were rigged even went on to say that the elections are rigged too, by the ruling party.

Questions on Exit Poll Prediction

Many people have asked me ‘how much trust do I have on these exit polls?’ For them this was a rather simple question, but for me this question is very difficult. How can I say something about the reliability of these polls without even knowing the detailed methodology about them? But even without looking at the details of each and every poll separately, and without looking at the numbers of each state, the one answer which I could give with reasonable degree of confidence was ‘the trends are very clear, BJP and NDA is heading for a very comfortable victory’.

The follow-up question often asked is, ‘do I see NDA heading for 400+ or is it close to 350+’. This is where it becomes even more difficult-- as in my opinion, in some exit polls I tend to agree with the overall figures but not the state numbers. Alternatively, for some exit polls, I may not agree with the overall figures, but I tend to believe figures for some states are correct.

For instance, I would find it hard to believe that in Tamil Nadu the BJP may be winning seats in the double digits; I would question the poll which indicated 4 or 5 seats for BJP in Kerala; though if the same poll suggests 9 or 10 seats for the BJP in Telangana, I would agree.

Differing Estimates for Nation and States

There are enormous variations to consider in an exit poll not only for the national tally but also for state-wise estimates. There are variations about Uttar Pradesh ranging from 60 to 72 seats, in West Bengal the consensus is that BJP’s tally will go up but the variation is from 22-23 seats to as high as 30 seats. I would find it difficult to believe that the NDA is performing as well as it did in 2019 in Bihar as there seems to be some changes on the ground. I can name various states where there are enormous variations in estimates of seats, Gujarat being an exception where all exit poll seems to suggest, the BJP is winning all the 26 seats. Some of the variations between two polls is different despite the vote share estimate being the same. So overall it is extremely difficult to say which exit poll may come closer to the real picture on the counting day.

Conflict in Recent Exit Poll Numbers

In the recent elections, many exit polls have proven erratic by throwing up conflicting results. Last year, many polls estimated an incorrect winner for Chhattisgarh and Madhya Pradesh, some exit polls were proven wrong in Rajasthan as well. There is no pattern to this, an agency may have predicted MP assembly election in the right direction, but made a mistake in estimating the results for Rajasthan.

There were states where all exit polls were correct (Telangana) and states where everyone was wrong (Chhattisgarh). In such a situation, how do you read these exit polls which released Saturday evening? How do you judge their accuracy? The memory of the 2004 Lok Sabha elections’ exit poll is still fresh in the mind of pollsters, analysts and politicians, when the consensus was that Atal Bihari Vajpayee-led BJP government was winning, but the results were very different. This was an occasion when all exit polls failed. Some believe (wishful thinking may be) that 2024 exit poll will also see the same fate as of 2004 exit polls. There have been many similarities between the contest of 2004 and 2024, at the moment, it is difficult for me to say if 2004 will be repeated in 2024 as far as the fate of exit polls are concerned, but in my understanding, it is an extremely distinct possibility.

I think it is important to look at the vote share estimates of the polls, because that gives us the best sense of support base of political parties which gets translated into winnable seats. Questions will be raised about the quality or reliability of such exit polls if they don’t estimate the vote share. It is similar to a doctor prescribing medicine without even asking the problem.