In a democracy, elections and the predictions and surveys associated with them are not a recent phenomenon. A look at the history of exit polls in India and abroad tells us that the Indian Institute of Public Opinion conducted the first election poll in India, during the second general election conducted in 1957. The Survey was headed by Eric de Costa; however, one might say that this was not a full-fledged exit poll.

Later, in 1980, Dr Prannoy Roy and Ashok Lahiri conducted the first exit poll for India Today magazine, which they repeated during the 1984 election. In 1996, an exit poll was conducted by Doordarshan, by journalist Nalini Singh, with the fieldwork and data gathering carried out by Centre for the Study of Developing Societies (CSDS).

In the US, the first exit poll took place in 1936. It was conducted by George Gallup and Claude Robinson in New York City, where they asked voters exiting the polling booths which presidential candidate they had voted for. Through an analysis of this data, they estimated that Franklin D. Roosevelt would win the election, which he did. After this, exit polls became popular in other countries too. Britain had its first exit poll in 1937 and France in 1938.

In India, the advent of television and later the internet has led to a shift in the purpose of exit polls. Once serving the sole purpose of research, the statistics are now primarily used as a means of electoral forecasting and seat count prediction. It is also true that many exit polls have been inaccurate in the past. Who can forget the 2004 Lok Sabha election? With the increasing hype and the metamorphosis of exit polls from a research-based study into a form of entertainment, the pressure for exit polls to gain popularity has also increased. With several exit polls turning out to be wildly incorrect in recent years, questions are now being about their credibility. There have been charges that exit polls are sponsored by specific interested parties or used for financial gains such as booking profits in the stock market. Amidst this, another point is raised—what is the need for exit polls when the election results are released just a few days later?

Why are exit polls important in a democracy

In a democracy, after votes have been cast, the most accurate way to understand the issues on which people voted for or rejected a candidate is through exit polls. Often, for instance, we see superficial issues dominating the campaign phase of an election, but exit polls reveal that people’s primary concerns are actually basic needs. Election results only tell us the outcome, not the reasons behind it. Exit polls, therefore, can gauge the pulse of a society by providing essential data-based information for further social science research into democratic institutions.

Also Read | Raising the EVM bogey

The significant inaccuracies in exit polls can stem from various reasons. Some election analysts suggest that exit polls often fail to accurately understand the opinions of the general public. Others believe that people are hesitant to provide honest opinions to surveyors. If public opinion cannot be accurately gauged, it can become a social science issue. If the same primary research method is used in other socioeconomic surveys, it raises questions about the credibility of those surveys as well. For example, if an agency’s exit poll is proven incorrect, it becomes apparent on election result day. But if the survey is a social study survey, such as the ASER one measuring the state of education, there is no immediate way to assess its reliability. If survey methodologies are flawed, it can create chronic problems for long-term policymaking and implementation.

Exploring why exit polls have been incorrect

Exit polls estimate vote shares from a random sample of voters and then attempt to convert this to seat predictions. It is challenging to predict national outcomes from small samples, especially when voters are reluctant to reveal their true choices.

In an ideal situation, the composition of a survey sample should accurately reflect the demographics of the community it represents. For example, if an electoral region includes 15 per cent minorities, 20 per cent Scheduled Castes/Scheduled Tribes, and 35 per cent OBC, then the survey sample from this area should have these groups represented in the same proportions. Additionally, variables like gender and urban versus rural populations should be carefully factored in to ensure balanced representation. However, surveys sometimes prioritise sample size over critical elements like representation and randomness, which can compromise accuracy.

Generally, the dominant section of society is overrepresented in samples (for example, the BJP in the 2024 Lok Sabha election or the Congress in the Haryana assembly election), while marginalised sections are underrepresented. Such misrepresentation is not intentional but occurs due to the social structure within which the research is conducted, with dominant groups more vocal and marginalised groups often silent voters. Consequently, qualitative underreporting occurs, meaning pollsters often miss predicting the swing votes of marginalised groups.

Due to the dominance of a few groups in the political discourse used to compile exit poll data, undecided responses such as “I don’t know” or “can’t say” are often associated with the stronger party, which increases the chances of inaccurate results.

Another critical aspect is the composition of surveyors in the field and whether they represent society proportionally. In some cases, surveyors are urban, college-educated, upper-class, or male. A more inclusive team (including women, rural dwellers, and people from marginalised communities) would find it easier to conduct social surveys.

In India’s first-past-the-post system, it becomes highly challenging to predict close contests accurately in exit polls. In addition, after the advent of the digital world, privacy concerns have emerged, making voters hesitant to reveal their choices. In the first-past-the-post system, the primary need is to capture vote share accurately. Converting vote share into seats is a separate exercise, and not a part of survey science. That is why in countries like Britain, only the vote share is reported and not the seat count. Agencies should clearly state that these numbers are accurate in terms of vote share, issues, and popularity, but that they may not be reflected vote share is converted into seat counts. Unfortunately, every exit poll agency insists on reporting seat share.

Need for improvement

Improving the accuracy and transparency of exit polls requires an overhaul of both regulatory measures and research methodologies. Exit polls are governed under Section 126A of the Representation of the People Act, 1951, with specific guidelines issued by the Election Commission of India to prevent them from unfairly influencing elections. However, these regulations could be expanded and modernised to address a range of challenges posed by variations in polling methods, sampling techniques, and technological advancements.

The diversity of methodologies used by different survey agencies is a significant issue. Some agencies conduct phone surveys while others gather data through in-person fieldwork. These differences can produce highly varied results, highlighting the need for a more standardised and transparent methodological process. Making the methodology, the sampling strategies, and the question designs publicly available would allow for a clearer understanding of the strengths and limitations of each survey. Transparency in methodology not only enhances trust but also promotes accountability, as voters and political analysts can better understand how the predictions are derived.

Further, just as businesses and services require licenses, it would be beneficial to mandate exit poll companies to undergo a registration and regulation process, supervised by the Election Commission or a Central government body. Licensing would help ensure that only qualified and unbiased agencies conduct exit polls while reducing the potential for manipulative practices and increasing data reliability.

Also Read | For fair representation

For greater transparency, exit poll agencies should also disclose critical information such as ownership details, the nature and sources of sponsorship, voter demographic profiles, and the formula by which vote share data is converted into seat projections. This will help the public better understand how the predictions are generated, and survey agencies can be held to a higher standard of accountability.

India could follow the examples of other nations. Many European and American countries have strict regulations on the timing of exit poll broadcasts, which helps prevent undue influence on undecided voters. As the world’s largest democracy, India could adopt similar measures. Although the Election Commission has the authority under Article 324 to regulate elections, additional Press Council or News Broadcasting Association rules could help create a consistent and reliable standard for exit polls across media platforms.

There is also a pressing need for survey agencies to adopt newer and more precise methodologies. By focusing on accuracy and methodological rigour and not yielding to pressures of popularity or profit, agencies can fulfil their primary responsibility, which is to provide data-driven insights into the electorate’s preferences without compromising on scientific validity.

Ashish Ranjan is an election researcher and co-founder of the Data Action Lab for Emerging Societies (DALES). Atul Kumar Pandey is an election researcher.