Exit Polls, Please Exit ASAP!

Gaargi Jamkar

TY BSc (2022-25)

Estimated Reading Time: 7 mins

Exit polls predicting Indian election results is like the IMD predicting the Indian Monsoons: they’re probably wrong!

Last month’s election results have been a surprise for everyone no matter their political alignment. A major problem factoring into that surprise was the exit polls indicating a sweeping NDA victory. The stock market had skyrocketed, investors were confident, and a BJP majority seemed obvious. However, on the day of the result, the image of a scenic landslide victory was painted over by a caricature of a coalition that somehow stayed intact due to what seemed like the grace of the almighty. Needless to say the stock market plummeted like they show in the movies and the atmosphere was strikingly similar to the ‘mood swings of a pubescent teen’ (all over the place). News channels were festive with the ‘plot twist’ and the memes flooded the internet.

My question here is, would there be any difference in the aftermath if the exit polls were more accurate or for instance, not been held at all?

Would one be less surprised?

Yes…definitely

Would the stock market’s volatility be cushioned to a certain extent?

Sure.

Would it have led to less heartbreak for Mr Modi?

Maybe? Who knows!

My point is that the exit polls showed us a picture that was far away from reality and the results came as a shock to most people whose expectations were completely blindsided. Had the exit polls been more accurate, the stock market wouldn’t have betrayed us and nor would the results be such a shocker to the citizens. Naturally, the next question that arises is how exactly are these exit polls predicted and why do they get everything wrong (especially in our country)?

source

Exit polls in India are conducted on the day of the election outside polling booths where the voters (citizens) are interviewed or are given a questionnaire to fill up anonymously. Now when it comes to prediction, there are many things any statistician should keep in mind: sample size, margin of error and behavioural biases. The exit polls don’t do a particularly good job of covering all these bases

Sample size: We all know that the best sample size is the population but we also know that achieving the same is impossible. The next best solution is to maximise the sample size as much as one can, to accurately represent the entire population. According to the Election Commission of India (ECI), the voting population for the 2024 election was close to 97 crores! So what is the ideal sample size then? 2 crore? 10 crore? 20 crore? No one knows exactly. This number is very difficult to identify and more likely than not, it is a large one. However, a larger number is not feasible for exit polls due to the time constraints and limited funds. Usually exit polls have a sample size between 5 to 10 lakh voters but such a number is not enough to comment on how many seats will be won by a particular party for a 97 crore electorate. The sample size isn’t just a number that needs to be tackled but also the ‘type’ of people that are interviewed matter. India is a diverse country with numerous parties in each state and people having different religions, backgrounds, castes, beliefs etc that are reflected in their vote. The sample size must therefore give proportional weightage to all ‘types’ of people. Such sampling is called ‘proportionate sampling’ that considers qualitative differences between people and accounts for it. Many exit polls in the country fail to achieve such a sample that reflects the diversity and fall prey to ‘convenience sampling’ or ‘accidental sampling’. Such sampling involves interviewing people by chance or without planning beforehand. Voters are selected based on the exit poll agency’s convenience and feasibility. How would such samples show the true picture of the entire country’s opinions?

Margin of error: As students of economics, we are told to validate each statement under a particular level of significance (usually 5%) to maximise the accuracy and avoid the chances of errors. Every number in a study is checked for statistical significance rather than numerical or magnitudinal significance. This however does not seem to be the case with exit polls. The numbers are released on TV channels but their margin of error is never reported. The predictions are made and carefully analysed but error margins are not taken seriously. This is especially problematic when one is already compromising on the sample size. It is also important to note that the exit polls not only predict the winner but go into a detailed analysis of how many seats will be won in which state and by what alliance. This requires a lot of data and time for analysis. Exit polls can only be released after all phases of the election (The Representation of the People’s Act, 1951) so they are short on both time and data (sample size is small) and yet the predictions are made in detail with lightning speed. Such discrepancies prove that exit polls are not a prediction, rather, a presumption.

Behavioural Bias: The most obvious and strong critic of the exit polls is the presence of the ‘social desirability bias’. Exit polls involve people self reporting the vote they caste in either interviews or questionnaires. Most people fall prey to this bias and lie about who they voted for. This could be due to fear of repercussions, fear of judgement or the need to conform with the rest of the population. People argue that keeping anonymity in interviews and questionnaires can eliminate the bias but there is no way of knowing whether the voter is lying or not. The law gives the right of secret ballot to every voter of the country. My question is, if one is unsure about the veracity of their data, why go ahead with the analysis and the so-called predictions?

Some countries like the UK have a different method for conducting exit polls. Here, a booth is kept outside polling stations and once a person has voted, they have the option to mimic the action by casting a vote at the exit poll booth. This method tries to replicate the actual voting procedure to get a more accurate dataset. This method has had fruition with the last several exit polls predicting roughly the same amount of seats for the parties as the results.

source

However, it is also important to note that Britain’s exit polls are one of the most expensive single question surveys in the world. Keep in mind that countries like Britain, France, Germany, Singapore, Korea, etc have smaller populations than our country. The diversity in people (religion, language, culture, etc) and even the number of parties competing is far less than that in India. This makes the population in these countries much more uniform in terms of voting preferences and easier to analyse. India being one of the most heterogenous and populated countries with numerous political parties in different regions faces a complex statistical problem when it comes to election predictions. This in itself raises the margin of error and gives less validity to the exit polls. The 2024 exit polls are not the exception, the 2004 exit polls predicted a BJP victory while the 2014 exit polls grossly underestimated the NDA alliance. This problem is not just limited to national exit polls where the sample size is impossible to account for, but also state wise assembly elections. The 2017 Uttar Pradesh or 2015 Bihar and Delhi exit polls egregiously miscalculated the results and were far from the actual results.

The point I want to make is that exit polls are pointless and futile. They are expensive to administer, the sample size is not large enough and there have been repeated failures since the past few decades. Most exit polls are funded or organised by News corporations like The Times Group, CVoter, India Today etc. These organisations conduct the polls so that they have something to air on the television and print in the news. These channels compete with each other and try to release their respective exit polls before the others. This leads to compromise in the quality of data and inaccurate predictions. The primary incentive to conduct exit polls is not citizen interest but to maximise TRPs (television ratings). So why should one believe what the exit polls are saying? There is little to none transparency in the procedure and methodology of the exit polls. Sampling techniques are not revealed, error margins are not reported and yet seat by seat analysis is shown, backed by so-called ‘statistical methods’.

To put it better into perspective, let me help with an analogy. Before the release of any movie, the production house releases a trailer. Think of the elections as a trailer to the release of the election results. Based on the trailer performance and attention, people speculate on whether the movie will be a hit or a flop. These speculations are not based on any actual data and are just conjecture. What matters is the box office collection on the opening day. The exit polls are the speculations that have no actual relation with the result of the elections. One could even argue that these speculations are done just to generate publicity (the same way exit polls are released to increase TV channel viewership). The election results are what matters in the end and the exit polls are immediately forgotten until next time. I guess this means that the ball is in your court; you can choose to google about the actor controversies and critic reviews or just sit back with a bucket of popcorn and enjoy the climax at the premiere!