Exit Polls Got it Wrong Because They Ignored India's Complex Ground Realities

With the Sensex recording a crash of 5500+ points, the faulty polls could be blamed for the bloodbath in the market.

(Photo: Vibhushita Singh/The Quint)

With nearly two-thirds of the votes counted, the NDA (National Democratic Alliance) is leading in 292 seats and the INDIA (Indian National Development Inclusive Alliance) bloc is leading in 235, with vote shares of 44.8 percent and 41.5 percent, respectively.

The BJP, individually, is at 238 seats and the Congress is at 100.

Some of the western and northern states sprung a surprise for the incumbent BJP, which lacked a clear national election narrative, while facing concerns of significant economic distress (as discussed here) in rural areas where the opposition has drastically gained both seats and vote share against the BJP.

See how Uttar Pradesh’s 80 seats saw a split outcome on the seat share of 39 for the NDA and 40 for the INDIA bloc, while Rajasthan’s 25 seats saw 14 for the BJP and 11 for the INDIA bloc.

In UP’s vote share data, one saw BJP+ at 44.12 percent (BJP: 42.13 percent), INDIA at 42.7 percent (SP: 31.98 percent, INC: 10.64 percent), and BSP at 9.2 percent.

In urban seats, the NDA's vote share this time is 52.8 percent, whereas in 2019, it was 48 percent.

But in semi-urban, semi-rural, and rural seats, the vote share accounts up to 39.9 percent vs 42.5 percent, 37.3 percent vs 42 percent, and 44.1 percent vs 44.9 percent, respectively, in comparison to 2019. It’s the mofussil towns where the BJP has had a fall (more details here).

The mandate of the Indian electorate with the gains made by the Congress and the INDIA bloc, gives a clear indication of a strong anti-incumbency wave, drawn from shared sentiments of increasing economic resentment and electoral disappointment against the incumbent, which was evident on the ground both against Modi and the BJP.

Economic issues of persistent price-rise (inflation), jobless growth, slow private investment, real wage stagnation, widening wealth-income inequality, poor economic incentive-led schemes, like PLI, at the cost of decreasing allocations for essential social welfare programs, like the MGNREGA and Nutrition Programs, etc, are all factors that have most people (especially in rural areas, which in the north, west, and parts of the south) voted heavily against the BJP.

It is likely that these rural and semi-urban electorate clusters will continue to be deeply impacted by these issues in the future too.

What is also evident now is the degree to which the exit polls got their predictions wrong by ‘predicting a landslide win for the BJP and the NDA’.

Thereby, allowing short-sellers in the stock market to make profitable gains (drawn from exit poll predictions), only to realise today while witnessing one of the largest falls in both Sensex and Nifty stock-prices (in recent months).

Sensex went down more than 5,500 points, and Nifty’s performance is illustrated below.

The intra-day stock market collapse raises serious questions about the extent to which market traders profited from a recent spike in stock trends’ established by the ‘irrational exuberance’ of ill-conceived, poorly designed, biased projections of exit polls (favouring a massive BJP win). 

There is a need to seriously question the design and methodological framework of how ‘exit polls’ were created, almost to favour or create a false pretence of a BJP landslide win. Maybe those who were aware of ground realities also operated under a fear psychosis to not upset the ‘incumbent party leaders’, which says a lot about the nature of mainstream media capture and the intellectual insincerity observed in the Modi era.

Methodological Issues: A Clear Lack of Transparency

Several major exit polls employed various methodologies to gather and analyse voter data. There were key issues with them.

For instance:

• News18 Mega Exit Poll: Conducted across 21 states and 518 seats, with over 95,000 voters surveyed, this poll utilised random sampling in the Vidhan Sabha constituencies, door-to-door interviews, and interviews at polling stations. They ensured quality control through geo-tagging and telephonic back-checks. Weak random sampling can give extremely biased  and asymmetric results.

• India Today-Axis My India Exit Poll: This poll involved interviews with voters exiting polling booths across multiple states using, again, random sampling techniques and applying statistical models for seat and vote share estimates.

• ABP News-C Voter Exit Poll: This poll is known for face-to-face interviews at polling stations, employing random and stratified sampling methods, and using proprietary statistical models. 

• Republic TV-P Marq Exit Poll: This poll conducted interviews with voters exiting polling booths, using random and stratified sampling techniques, and applying vote-swing calculations.

While these methodologies sound robust, a crucial gap remains: most exit polls do not explicitly mention the confidence levels, margins of error, or even the size of their samples which, in any case, would be microscopic compared to the electorate size of around 100 crore. 

This complete lack of transparency, or proactive disclosure, raises questions about the reliability of their projections.

Knowing the sample size, understanding the confidence level, and the margin of error is essential in gauging the accuracy and precision of any statistical prediction, especially in elections.

Exit polls usually provide aggregated figures for states or the entire nation, which can be misleading. While they might achieve high accuracy in macro results, they often fail at the micro level, particularly in individual constituencies. This discrepancy underscores the necessity for detailed projections at the constituency level to provide a clearer picture.

In other words, the true test of these exit polls would be their constituency-wise projection of the winners.

Otherwise, a correct prediction might be akin to reaching the right answer in a school math class but getting the entire working of the answer or process of arriving at an answer wrong.

For poll-astrologers, perhaps it is more transparent to stick their necks out and present their predicted winners for all the 542 (543-1) Lok Sabha constituencies. Anything less than that, in any informed observer’s humble opinion, is just crystal gazing, number crunching, or juggling.

In these surprising elections, Modi and the BJP may have also appeared to discover the limits of hype in Indian politics.

“An apparently unified public sphere, solidly pro-government media, and impressive growth numbers had left many assuming that Modi’s performance in power had few holes.”

This author cautioned against observing a waning effect in Modi’s popularity this election, which seems to be clearly established now, particularly in rural and semi-urban areas.

The Indian electorate is also quite conscious of what’s affecting them and, more importantly, isn’t influenced by hype, propaganda, and statistical jugglery, designed to project a story around growth that hasn’t reached the electorate’s wallets. 

On exit-polls, bias, and the hype created around them, it must be acknowledged that the current pollsters’ practice of not disclosing complete methodological details, particularly the confidence levels and margins of error, diminishes their credibility and validity. The effects are dire, as seen from what’s happened to India’s stock market performance in the past few days.

Until exit polls provide detailed projections at the constituency level and disclose all relevant statistical parameters, their reliability will remain questionable.

As the actual election results unfold, the shortcomings of these exit polls become evident, highlighting the need for greater transparency and methodological rigour in future polling exercises.

(Deepanshu Mohan is Professor of Economics, Dean, IDEAS, Office of Inter Disciplinary Studies, and Director, Centre for New Economics Studies (CNES), OP Jindal Global University. He is a Visiting Professor at the London School of Economics, and a 2024 Fall Academic Visitor to the Faculty of Asian and Middle Eastern Studies, University of Oxford. This is an opinion piece and the views expressed above are the author’s own. The Quint neither endorses, nor is responsible for them.)