Bihar election results didn’t just see-saw on Tuesday. Most opinion polls before the elections gave the BJP-led NDA an edge over the RJD-led Mahagathbandgan. By the time the elections were over, almost all exit polls were showing the Mahagathbandan would win. On the day of the results, the exit polls seemed to be proving correct, before the pendulum swung back and gave the NDA just the numbers needed for a tiny majority.
While polling has been called an “inexact science”, especially in a complex and diverse country like India, it is uncommon, though not unheard of, for almost all pollsters to not just get the number of seats for the winning party wrong but also the entire direction of the results.
Two polling agencies — India Today-Axis My India, and CNN-News 18-Today’s Chanakya —which have had a good streak the last few times were the farthest from the results this time.
Exit polls of both predicted landslide victories for the Mahagathbandhan.
ABP News-CVoter, Times Now-CVoter, Republic-Jan Ki Baat, and India Ahead-ETG gave the RJD-led alliance an edge.
A day after the results, a founder of one of the polling agencies said they were “still evaluating what went wrong”. Two others said the reason was that they did not read the women voters right.
In a statement on Wednesday, Axis My India said that “overall female voter turnout was 5% more than male… which swung the polls in NDA’s favour”, with the women also voting 5% “more in NDA’s favour”.
It added that while in the first stage of the three-phased elections, “men voted 1% more than women”, women were up by 6% and 11%, respectively, in the next two. The agency stated that their predictions were “exactly in line with results of the first phase but went amiss for the second and third phase” and claimed that “Covid-19 restrictions limited our number of women voter interviews”.
Yashvant Deshmukh, who runs CVoter International that claims to do a daily tracker, said most of the pollsters got it wrong because they “only tapped the vocal voters”, women being one of the silent ones. His guess was that while men voted for the Mahagathbandhan, the women preferred the NDA. CVoter did exit polls for two different news channels.
There was another, more fundamental reason for the predictions seeming so off the mark, according to Deshmukh. “In India media and editors are focused on seat share positions… The science of survey starts and stops at vote share. (The seat share is extrapolated from the vote share.)… Very few journalists in India understand what to expect from the polls… We don’t have an appetite for data.”
Deshmukh went on to point out that CVoter’s exit poll data got the vote share right. It said the NDA would get 37.7% of the votes, and it got 37.4%.; for the Mahagathbandan, it predicted 36.3% votes, it got 37%. “It was too close to call (many seats were decided by thin margins). But if you get the vote share wrong, you are going to get the wrong seat projection as well.”
At the same time, Deshmukh admitted, there have been instances where pollsters have got the vote share wrong and still somehow got the winning seat share. However, he said, that didn’t make the pollster right.
Off-record, most pollsters said that beyond the seat numbers, the polls should be read for trends.
📣 The Indian Express is now on Telegram. Click here to join our channel (@indianexpress) and stay updated with the latest headlines