Opinion polls

The prediction industry

Print edition : June 10, 2016

PREDICTING election results in Tamil Nadu, particularly in the case of the Assembly, is a hazardous exercise. The election prediction industry, which works with surveys and opinion polls, has come a cropper on more than one occasion. The 2016 Assembly election in the State will stand out for one other reason: the deluge of opinion polls spewing out numbers on the television channels that smelt an opportunity to increase their rating points.

The perils of predicting election results are many in a State like Tamil Nadu, which has a staggering 80 per cent of its population on the electoral rolls and a multifarious sociocultural and political profile shaped by the Dravidian movement and other identity-based political formations at different time periods. Political mind-reading here involves a laborious exercise of intricately fusing both the past and present and using the tools of science and technology to unravel the outcome.

The 2016 elections had an added element in the form of the People’s Welfare Front (PWF) to challenge the traditional dominance of the two Dravidian parties—the All India Anna Dravida Munnetra Kazhagam (AIADMK) and the Dravida Munnetra Kazhagam (DMK). The partners in the PWF—the Marumalarchi Dravida Munnetra Kazhagam (MDMK), the Viduthalai Chiruthaigal Katchi (VCK), the Communist Party of India (CPI), the CPI(Marxist)—and the Desiya Murpokku Dravida Kazhagam (DMDK) of Vijayakanth and the Tamil Maanila Congress (TMC) led by G.K. Vasan, which also went with it, had formed alliances with either of the two Dravidian parties in previous outings to set up bipolar contests.

The majority of the opinion polls and exit polls this time favoured the DMK-Congress combine. But a few, such as Thanthi TV-Krish Info Media, Times Now-CVoter and News Nation, predicted a return of the AIADMK, and their exit polls also reaffirmed the results of their opinion polls.

Thanthi TV’s opinion poll was an expansive exercise and had representative samples (150 in each Assembly segment adding to 35,600 for the State as a whole) from all constituencies. It said the AIADMK would win 102 seats and the DMK 79. Its exit poll gave the AIADMK 111 and the DMK 99. They were not able to predict the results in 16 seats, which were too close to call. The remaining 24 seats were distributed among the Pattali Makkal Katchi (PMK), the PWF and the Bharatiya Janata Party (BJP). “Our methodology was exact. Our error margin also worked well. That is why we were able to predict close to the final result,” said Arun Krishnamurthy, founder of Krish Info Media. Krishnamurthy told Frontline that the company’s methodology was unique. It devised a strategy known as “mood mapping”, in which the idea was to understand the voter’s mindset by eliciting the voter’s views on various issues and not confining the whole exercise to just his voting preferences. “The questionnaire was carefully drafted, with sections to specifically understand the electorate’s choices and opinions on key issues. We gave ample time to respondents. This helped enormously to arrive at a decision on their personal choice to be made in the election,” he said.

However, Krishnamurthy and his team, which made a close prediction in the 2014 parliamentary election, again for Thanthi TV, said that this time they had carefully chosen their sample by giving weightage to all factors such as caste, gender, religion, language and region as per the records available with the Election Commission of India (ECI).

“This again helped us to give a representational sample within the random order and also enabled us to assess the voting trend. We clearly divided every constituency into separate pockets, on the basis of which we designed the poll samples,” he further said.

The other survey that predicted an AIADMK win was done by the Tamil TV channel Puthiya Thalaimurai, in association with an agency called APT. However, its prediction was way off the mark. It suggested that the party would get 164 seats with a vote share of 38.58 per cent and that the DMK would come a distant second with 66 seats and a vote share of 32.11 per cent. The sample size was 4,999 persons across Tamil Nadu, and the polling was done between April 18 and May 4. The result was released just a week prior to May 16, the election day. The survey further said that the PMK would win 15 per cent of the votes from the northern region alone.

The Times Now-CVoter poll conducted in March 2016 made a near-correct prediction by giving the AIADMK an absolute majority of 130 seats with a vote share of 39 per cent, followed by 70 for the DMK combine with 32 per cent of the votes. The CVoter’s other survey with India TV in February 2016 said it was a tough fight between the AIADMK (116) and the DMK (101), with no party getting a majority. The News Nation said in March that the DMK would win between 107 and 111 seats, and the AIADMK between 103 and 107. NDTV’s survey gave the DMK a comfortable lead and stated that since 1984 there had been an average swing of 10 per cent against the incumbent.

The Chennai edition of the popular Tamil daily Dinamalar in association with the Tamil TV channel News 7 conducted a survey with a sample size of 2.34 lakh people, or 1,000 per constituency. It predicted a clear majority for the DMK front. It revealed neither the methodology nor the breakdown of the sample. A source said the survey was conducted through college students, media persons and newspaper vendors.

A study conducted by the multidisciplinary Research Study Centre of People Studies of Loyola College in Chennai said the DMK had the edge. It took a sample size of 8,064 respondents spread across 177 constituencies and indicated that the DMK would garner 42.7 per cent of the total votes, followed by the AIADMK with 36.6 per cent.

In an earlier survey conducted between January 7 and January 19 by the centre, the DMK was sitting pretty. The survey was conducted in 39 parliamentary constituencies and 130 Assembly constituencies and consisted of a sample size of 5,464 people. Senior functionaries in the AIADMK, however, refuted it by saying that the sampling size was too small to evaluate the vote share accurately. Junior Vikatan, a Tamil magazine, said that the DMK alliance would get 77 seats against the AIADMK’s 73. It said that it was not able to predict the results in 83 seats, as they would witness close contests. The survey gave the PMK a lone win.

As for exit polls, barring the Times Now-CVoter and Thanthi TV polls, the majority of the others gave the DMK the edge. While Times Now-CVoter stuck to its opinion poll findings of the AIADMK gaining 130 seats, Thanthi TV predicted 111 for the AIADMK and 99 for the DMK. It also declared that 16 constituencies were too tough to make a prediction.

Among the others that carried out surveys were the RG Flash Team, Cauvery News, Goodwill Communications, the magazines Thuglak, Kumudam Reporter and Nakkeeran, and VDP Associates. Online surveys in social media were also carried out. Thus, the conflicting results of such multiple surveys, which many claimed would influence voters, had opened up a debate on whether these surveys should be allowed during or prior to elections.

The ECI, in its notification under Sub-section 1 and 2 of the Section 126A of the Representation of the People Act, 1951, prohibited the publishing of results of any exit poll in connection with the elections to the Assemblies of Assam, Kerala, Tamil Nadu, West Bengal and Puducherry during the period between 7 a.m. on April 4 and 6.30 p.m. on May 16. However, it did not prohibit pre-election surveys.

Complex situation

What made this round of election a psephologist’s nightmare? Alliance arithmetic has been a major factor in Tamil Nadu elections. But this time, neither of the Dravidian parties forged any grand alliance, although the DMK took the Congress (41 seats) and a few smaller parties into its fold. But a reasonably healthy support to both these parties and the absence of a wave of resentment against the ruling dispensation had made the reading a complex one. Adding to the complexity of the situation was the presence of a six-party alliance, the PMK, the BJP and the Naam Tamilar Katchi claiming to provide a real alternative to the two Dravidian parties.

According to the Election Commission, the general electorate in Tamil Nadu this time included 1.38 crore young voters in the 18-29 age group, constituting about 23 per cent of the total electorate of 5.82 crore compared with 15 per cent in 2011.

The other cluster was the first-time voters in the 18-19 age group, who swelled from 8.56 lakh in 2011 to 21.05 lakh in 2016. These two clusters could be classified as “neutral” and could have voted for any party.

“It is a difficult proposition to read the mind of these two youth clusters, which have emerged as a significant force in this election. Also, many did not vote,” said a political analyst. Besides, the voters could also opt for “strategic” voting, i.e., taking a last-minute decision, thus upsetting any study.

Cash for votes

But there was one more factor vitiating the elections this time in the State—cash for votes. In any close battle, such an extraneous factor is bound to play a significant role in the outcome, denying other players a level playing field. “I was confident. Just a day prior to the election date, everything changed. Money was raining. Money defeated us,” said VCK leader and candidate in Vanur, D. Ravikumar. He finished fourth.

But whether the swathe of pre- and post-election surveys in various television media and print media are based on scientifically designed and technologically driven methodologies or on naive presumptions is a million-dollar question today. Many of these studies, academicians claimed, are opinionated and for television ratings since the media in Tamil Nadu is heavily politicised and highly competitive. Nobody knows how they arrived at the figures and conclusions.

These experts also claimed that in such predictions, sampling is the key factor. To achieve this, the voter’s social, cultural and political profile in each constituency should be factored into the sampling selection. Besides, factors such as past trends, polling patterns, political affiliations, alliance arithmetic, caste, religion, minorities, money flow, candidate and constituency profile, all need to be taken into account.

“Unscientific selection of sampling often fails to reflect the exact mood. The sample size should be quantitative, qualitative and diversified. A foolproof formula to assess the conversion of vote share into seats has to be employed,” Ravikumar said. The noted psephologist Yogendra Yadav said Tamil Nadu elections had been hard to predict. He was surprised with two polls in the past going exactly the opposite way. The Danish physicist Niels Bohr once said: “Prediction is very difficult, especially about the future.” And Tamil Nadu elections are no exception.

Ilangovan Rajasekaran

This article is closed for comments.
Please Email the Editor