Polling public opinion, choices or mood at any time in India is a hazardous business given the inadequacies of the Indian landscape and innumerable variables that must be factored in to arrive at a rationale and result. The risks are higher when it comes to exit polls, given the subjugation of facts by articles of faith. The fact that findings/projections of exit polls in recent years have gone awry has only made it more perilous.
It would be fair to say that exit polls are haunted by questions. The commentary on the accuracy and credibility of exit polls depends on the ‘who’ factor—entrenched emotions of identity and ideology rather than the matrix of mathematics. Typically, those winning are happy to accept the findings, and those losing choose to challenge the very concept of exit polls. The favourite illustration critics present for condemnation is the projections of the 2004 General Elections—indeed both opinion polls and exit polls blundered, perhaps blinded by the rhetoric of India Shining.
In 2004, the NDA was projected to win between 230 and 275 seats, and the UPA in the range of 176 and 190 seats. Actual results found the pollsters on the wrong side of both trajectory and tally—the NDA won 187 seats and UPA 219. The BJP won 138 seats and the Congress 145 seats and the call from Rashtrapati Bhavan. Five years later, the 2009 exit polls got the trajectory right but missed the tally—the projection for UPA was a low of 185 and a high of 205, and for NDA it was 165 to 195 seats. On results day, UPA won 262 seats and NDA got 159 seats.
The performance of pollsters in 2014 was similar—they got the mood of the voters but missed the tally by a wide margin. The exit polls gave the NDA between 183 and 289 seats, while the UPA tally was projected to be between 92 and 120. On May 16, 2014 NDA bagged 336 seats and the UPA tally plummeted to 60, with the BJP at 282 and the Congress at a historic low of 44 seats. The big news about 2014 was the BJP-led NDA sweeping 73 Lok Sabha seats in Uttar Pradesh. What did exit polls predict? Common consensus was between 49 and 54, with one outlier at 67. The SP and BSP were projected to win 11 and 15, whereas the final score read SP 5 and BSP 0.
The concept of exit polls used across the world is an innovation of Warren J Mitofsky. While at the US Census Bureau, Mitofsky designed a number of surveys. Along with Joseph Waksberg, Mitofsky devised a random digit dialling system for phone polling voters and went on to create the CBS News / New York Times poll. Success, Mitofsky underlined, required trained interviewers, an established pattern, and preciseness in calculations.
Exit polls have had a chequered history in India and elsewhere. Critics believe the underlying construct of exit polls challenges the definition of random sampling, and therefore the polls carry a higher margin of error. Like other polls, exit polls can overstate the case of vocal voters and miss the silent vote—and in India, there is an another factor, false responses driven by fear of retribution. Also, a higher turnout can skew assumed weightages, leading to erroneous calls on trajectory and/or tally. In fact, the impact is aggravated when the data is drilled to deliver outcomes at a granular level.
Indeed the track record of exit polls in Assembly elections validates what is seen as problematic. Take the 2015 Assembly polls in Delhi and Bihar. In Delhi, the polls sensed the trajectory but none of the pollsters could capture the sweep of 67 seats by AAP versus 3 for BJP and 0 for Congress. In Bihar, the exit polls got both trajectory and tally wrong. The consensus view was that the NDA had the edge and would bag between 111 and 155 seats. When the results came in, RJD-plus had won 155 and NDA 83 seats—the one poll which got both trajectory and tally withheld its findings. In Tamil Nadu, six of nine pollsters blundered on both trajectory and tally.
It is not as if exit polls have not got it right—for instance in Assam, West Bengal and Kerala. Mostly it has been the direction of the vote and sometimes a tally close enough to the actual results. Exit polls also got the mood of UP right in 2017, but could not capture the spectacular sweep by the BJP-led alliance, which won 325 seats. The pollsters got the trajectory but missed the tally—none predicted a tally of 300-plus.
By the nature of their construct, exit polls can get the direction right —for instance, in Gujarat, almost all the pollsters got the vote share of both BJP and Congress right. Exit polls also tend to get it right when there is a clear edge for one side at the outset of the election. On the flip side, exit polls can go haywire in close contests and when a thin sample is extrapolated to generate conclusions.
The crux of credibility is in what pollsters chose to do or not to do. For instance, the fog of X factor is triggered by the conversion of vote share into seats. Pollsters could enhance credibility by focussing on vote share and trajectory. That may be a tough ask in the TRP era when seat tally is an essential item number. Ergo, on Sunday, it would be a good idea to set the salt and pepper shaker along with the silverware as the projections are unveiled.
shankkar.aiyar@gmail.com