Exit polls will be in the dog house for some time to come

Nearly 2 decades ago, when the Hindustan Times was first launched in Mumbai, one of the cutting edge offerings it gave city readers was opinion polls in quick time ranging from subjects like: ‘Is the Brihanmumbai Municipal Corporation (BMC) to blame for the water shortage?’ to ‘Should Bollywood stars enter politics?’

The ‘vendor’ who did the phone interviews and crunched the numbers for HT was ‘C-Voter’. As coordinators for these quick surveys, we were told C-Voter was required to contact at least a 100 respondents keeping income and cultural backgrounds in view. Later, we unofficially learnt the sample base was one-tenth the target, and much of the projections were often ‘manufactured’.

Cut to the current exit poll blowout. The sight of ‘Axis My India’ pollster Pradeep Gupta weeping copiously on camera on the ‘India Today’ channel after his 361-401 seats prediction for the NDA went awry, will not be easily forgotten.

Shaky history

For years it is known most pollsters have been skating on thin ice. This year the issue has taken a more serious turn for two reasons. First, the charge by the Opposition parties that the poll projections were ‘scripted’ and not based on ground reports. Second, and more serious is many, including Congress’ Rahul Gandhi, have alleged pollsters of being in league with stock market punters to talk up the market before counting day, and then triggering a crash that wiped Rs 30 lakh crore off the capital markets.

Exit polls have got it wrong in the past, but this time the margin of error is a scandal. Most of them including India Today-Axis My India, Chanakya and ABP-CVoter, predicted the BJP would cross its 2019 mark of 303 and touch around 330 seats on its own while the that National Democratic Alliance (NDA) would get 350-400 seats. On the other hand, they said the Congress would get 52-67 seats and the INDIA Alliance would average about 140. The truth was the NDA got 293, the BJP 240, the Congress 100, and the INDIA Alliance 238. That’s about a difference of a 100 seats in each category and they are 25 to 35 percent off the mark!

Exit polls have gone horribly wrong in the past too. For instance, while the 2004 Lok Sabha was an upset win by the Congress and allies, exit polls predicted a comfortable ‘India Shining’ victory between 240 to 275 seats for the NDA. The reality: the BJP-led Alliance won only 187 seats, while the Congress and its allies bagged 216 seats, and formed the government.

Exit polls are not popular in the US or Britain. What they do there are opinion polls. Even then, pollsters like Gallup usually get their forecasts right allowing a maximum margin of error of 4 percent.

In the 2020 US Presidential elections, many pollsters were blamed for giving an excessive advantage to Joe Biden, but when the votes were counted the margin of difference with the forecasters was about 1%. The big poll organisations – Gallup, Reuters-Ipsos and the Quinnipiac University Poll – use a not-too-large sample size of 1,000, but the survey is scientific and respondents chosen are representative of the electorate. A maximum error of 3% is what is considered a successful forecast.