Elections
Exit polls 2024: Track record could tell us whom to trust
Summary
- Today’s Chanakya had the most accurate exit poll predictions in 2014 and 2019, while Axis My India also got it right in 2019. Mint's analysis shows exit poll forecasts need to be taken with a pinch of salt.
The largest dance of democracy came to an end on Saturday with the last phase of the Lok Sabha election. Exit polls have predicted a sweeping win for the ruling Bharatiya Janata Party-led National Democratic Alliance.
With so many predictions, the past record of each pollster may be the only way to assess how the final tally might turn out to be on Tuesday, 4 June, when the Election Commission of India releases the results.
In the last two Lok Sabha polls, very few prominent exit poll agencies were able to predict the results within a reasonably accurate range.
Exit polls are based on what a sample of voters tell surveyors as they exit the polling booth after casting their votes. Today's Chanakya, Ipsos, Axis My India, and CVoter are some of the agencies that have conducted exit polls over the past decade in collaboration with various media outlets, and some have given their estimates this year as well.
The list of pollsters keeps changing every year, and Mint have considered some select agencies whose predictions were available for the previous two national elections.
How exit polls fared in 2014 and 2019
At least two polling agencies—Today's Chanakya and Axis My India—made somewhat accurate predictions to varying degrees in the 2014 and 2019 Lok Sabha elections.
For the purpose of this analysis, the errors are based on a comparison of an alliance’s final seat tally and the midpoint of the range that a pollster predicted. For example, if a pollster predicted 300-350 seats for an alliance, which ended up winning 350 seats, the error is calculated as 25 out of 350, which is 7.1%. Pollsters typically have an acceptable tolerance range for the error, since predicting the exact figure is not possible.
Today's Chanakya had the most accurate prediction for the National Democratic Alliance in both the 2014 and 2019 national polls, with a minimal error of 0.8% in 2019, and 1.2% error in 2014.
The agency had predicted 350 seats for NDA in 2019, and 340 in 2014. The BJP-led alliance emerged victorious on both occasions, securing 353 and 336 seats, respectively.
Similarly, the agency’s forecast for the Congress-led United Progressive Alliance was also close to the mark. It predicted 95 seats in 2019, and 70 in 2014 for the UPA, which ended up garnering a total of 93 seats and 60 seats those two years, respectively.
When Axis My India and Ipsos forecast a grand ‘exit’ for UPA
Two other agencies—Axis My India and Ipsos—also saw their predictions hit the bull’s eye at least in one case each. The midpoint of Axis My India’s prediction for the UPA was 93 seats in 2019, which is exactly what the alliance won.
Ipsos’ prediction for the NDA in 2019 was 336 seats, close to the tally of 353.
Note that an agency's record cannot be judged just by Lok Sabha elections, since it may have reached different conclusions for state assembly elections.
CVoter and Nielsen way off the mark
The estimates by CVoter and Nielsen had much bigger errors. CVoter had an error of nearly 19% in its NDA estimate for 2014, and 14% in 2019. Nielsen’s error was close to 20% those two years. However, both agencies were right in predicting a win for the NDA.
The two agencies’ predictions for the UPA were even further off the mark. In 2019, CVoter predicted 128 seats for the UPA, which ended up securing just 93 seats. The error was bigger in 2014.
Nielsen’s prediction for the UPA was 97 and 130 seats in 2014 and 2019, respectively.
