Journalists try to explain political dynamics during elections through electoral arithmetic and electoral chemistry. While they rely on a range of statistics for the former, they try to gauge the popular mood of the people through field reporting to discern the latter. However, over the last two decades, opinion polls seem to have replaced conventional journalistic wisdom. Media houses, especially television channels, began giving primacy to surveys — both pre-poll and post-poll — to capture the political trend.
Relying on opinion polls to get an idea of which way the political wind is blowing has turned out to be a futile exercise since the ‘India Shining’ campaign of the National Democratic Alliance in 2004. Barring exceptions, most polls have got their numbers wrong. The astounding victory crafted by Prime Minister Narendra Modi and Bharatiya Janata Party president Amit Shah in Uttar Pradesh was not captured by any poll.
Getting it wrong
The Indian media is not alone in this conundrum called opinion polls. Most polls in the U.K. did not predict the Brexit result and the American polls failed to predict Donald Trump’s victory. In India, recent examples have been the failure to predict the victory of the All India Anna Dravida Munnetra Kazhagam (AIADMK) in Tamil Nadu in 2016 and the victory of the grand alliance of the Janata Dal (United), the Rashtriya Janata Dal and the Congress in Bihar in 2015. David Uberti of Columbia Journalism Review in an essay, ‘How polling data can be dangerous for political journalists’, explained the pitfalls in relying on opinion polls to gauge the political mood.
I had a humbling experience with pre-election opinion polls in 1998. I was working for Outlook magazine and we commissioned AC Nielsen to do the survey. We predicted a landslide victory for the Dravida Munnetra Kazhagam and Tamil Manila Congress front. Instead, the AIADMK-led front made a dramatic comeback. I had to write an article explaining our failure. Partha Rakshit, Managing Director of AC Nielsen, wrote a supplementary piece, ‘Hazards of Forecasting’, that talked about the problems in projections from the day on which the survey was conducted to the swing forward into voting day. I tend to agree with sociologist Herbert Gans: “Polls are not the best representative of the popular will, for people’s answers to pollster questions are not quite the same as their opinions — or, for that matter, public opinion.” The electoral outcome is an organic manifestation of the people’s will. An opinion poll is an inadequate tool to comprehend this democratic spirit.
The Pew Research Center published an article, ‘Why 2016 election polls missed their mark’, listing three major factors: non-response bias, many of those who were polled were simply not honest about whom they intended to vote for, and the way pollsters identify likely voters. The Pew Research Center explains non-response bias: “It is possible that the frustration and anti-institutional feelings that drove the Trump campaign may also have aligned with an unwillingness to respond to polls.” The article concedes that identifying likely voters is “is a notoriously difficult task”.
Many questions
The problem with journalism, which is akin to the social sciences, is that it wants to mimic the fundamental sciences. Why should it strive for a number? Is it not enough if it explains the processes, the players and the different possible outcomes? What is the place for guesstimates that masquerade as scientific predictions in daily journalism? Some questions for which I have not got convincing answers are: Is there an ideal sample size? Is there a proven formula to convert the vote share to the number of seats? Is it possible to extrapolate the vote share pattern of one region to a larger unit, say, a State or the country?
In this context, I am delighted with this newspaper’s decision to deploy its resources in extensive reportage and to not commission election surveys. Surveys have a speculative quality to them. Good reports, on the other hand, always capture a sliver of life.
