How accurate are exit polls? For starters, the so-called exit polls are largely ‘post-poll surveys’.
There is a technical difference between the two terms. Exit polls, strictly speaking, are done outside the polling stations immediately after voters come out. A post-poll survey, on the other hand, is undertaken a day or two after the voting, preferably at the homes of the voters.
Pollsters claim these surveys help capture the results better as they are held in the confines of somebody’s home.
Besides, they are based on relatively leisurely and detailed interactions, unlike exit polls where voters are caught immediately after emerging from the booths.
“The last time we did exit polls was in 1996, which was telecast on Doordarshan We got the national picture correct, but we were way off the mark for the state-level numbers,” Sanjay Kumar, Co-Director of Lokniti - Centre for Study of Developing Societies (CSDS), told Business Line .
More accurate According to Kumar, post-poll surveys are more accurate than exit polls. Lokniti-CSDS is doing a post-poll survey for CNN-IBN, just as NDTV has commissioned Hansa Research for its survey.
Both CNN-IBN and NDTV are releasing their final numbers only on Wednesday. “Part of the reason why exit polls have gone wrong in the past is too hasty an analysis. That’s why NDTV’s exit poll will only come on Wednesday,” tweeted Vikram Chandra, Group CEO, NDTV Networks.
On the other hand, the Times Now-ORG India says it has done a ‘mega’ exit poll. The CSDS-CNN IBN post-poll survey claimed to have a total sample size of 21,044 voters across 1,284 locations of 287 constituencies.
The Times Now exit poll claims to have ‘mapped’ over 5.4 lakh voters spread over 180 constituencies.
Immediate results
According to Ranjit Chib, a consultant with AC Neilsen, most agencies have migrated to post-poll surveys to benefit from the process, given that elections are mostly being done in phases. Exit polls are held only when there are single-phase elections as in the case of States such as Rajasthan.
Mixed polls “Agencies mainly use a mix of post-poll survey and exit polls on the last day of elections to generate immediate results, under pressure from television channels,” noted another pollster, who did not wish to be named.
According to Lokniti-CSDS’ Kumar, the desired level of ‘randomisation’ is lower in exit polls, where those coming out from the booths and willing to divulge their choice are more likely to be young urban males.
Women, people voting for smaller parties or belonging to lower sections, and those from deeper rural sections tend to get under-represented in the samples. “These issues get sorted out when we choose voters from the list spread across the various categories,” he added.