Exit polls close the gap with ads

Skybus ads across newspapers in Assam by major political parties raise serious questions about the business of polling

After the first phase of poll in Assam, a number of newspapers carried a skybus ad — an eight-column advertisement below the masthead of a newspaper — of the Bharatiya Janata Party (BJP) that read like a banner headline: “BJP to win all constituencies of Upper Assam”. Interestingly, although the Congress Party approached the Election Commission complaining about the deceptive advertisement, it eventually responded in a similar fashion.  

Just two days later, a similar skybus ad was published in a number of newspapers in the state: “Congress wave across Assam”. In both cases, however, it was clearly indicated on the left side of the ad that it was paid for and published by the respective political party. Still, some people expressed concern that these might look like news items or might sound like an exit poll prediction, at least to some electorates who might not be “smart” enough to understand that these were just advertisements. Whether such an ad amid the election should be allowed or not is certainly within the purview of the Election Commission. I, however, prefer to look at the other side of the coin.

One of the newspapers that published the ad, in fact, clarified: “No exit poll regarding election results was ever conducted by the newspaper, nor have results of such exit polls been published.” Well, was there any scope of misinterpreting these skybus ads as partial exit polls, at least? In other words, do the exit polls, in reality, sound similar to ads? Are exit polls self-fulfilling prophecies only?

While opinion polls are capable of influencing the voting behaviour in many ways, exit polls are more of fun, and its history is relatively recent. American pollster Warren Mitofsky and his friend, George Fine, introduced exit polling in 1967 by conducting interviews of voters as they left the polls in a gubernatorial contest in Kentucky, and exit polling had become a staple of elections nationwide in the US by 1972. Often systematic sampling is adopted for conducting exit polls where, say, every fifth or eighth voter of some booths may be interviewed. Any enthusiastic voter eager to opine in between should be discarded.

It’s not that exit polls always predicted correctly ever since its inception. The theory of “shy Torry voters” popped up when exit polls made a mistake to predict John Major’s victory in the 1992 UK general elections. And then in the 2000 US presidential elections, Voter News Service (VNS), a consortium of the major networks and the Associated Press, projected first Al Gore and then George W Bush the winner in the excruciatingly close and decisive state of Florida before pulling back in the middle of the night and declaring the contest too close to call.

Yes, it was really too close to call, and the credibility of exit polls was reasonably sound around that time. The Rose Revolution of Georgia in 2003, which led to the ouster of Eduard Shevardnadze, was strengthened by exit poll results. Venezuelan recall referendum of 2004 to determine whether Hugo Chávez should be recalled from office created massive uproar due to wide discrepancy with exit poll. Ukraine’s Orange Revolution of 2004-05 was also triggered by exit poll predictions. All around the same time. However, possibly the decline of the credibility of exit polls worldwide began around the same time as well.

People still struggle to explain the reason behind the 6.5 per cent discrepancy between the actual result and the exit poll in the margin of vote shares between George W Bush and John Kerry in the 2004 US presidential elections. The final national exit poll was considered to be the real “Smoking Gun” which even motivated an analyst like Richard Charnin to examine the possibility of phantom voters and uncounted votes in the election procedure. Some political pundits, however, asserted that the poll simply “behaved badly”. Thereafter exit polls failed miserably in many important elections worldwide. In 2004 Lok Sabha elections, the United Progressive Alliance (UPA) sprang a surprise by winning 218 seats against the prediction of 170-205 seats by most of the exit polls. While most polls had indeed predicted a National Democratic Alliance victory in 2014, most of them couldn’t predict 336 seats. Similarly, the 2015 Assembly elections of Delhi and Bihar, the 2016 Assembly elections in Tamil Nadu, the 2017 Assembly elections of Uttar Pradesh are examples of failures of exit polls — either to spot the eventual winner, or to estimate the quantum of victory of the winner.

What are the reasons for such debacles? Do all the pollsters conduct polls following proper statistical theory and rigour? Do they conduct their surveys in remote parts of the country and also in the sensitive booths as well with due importance? Understandably, in the rush to release their predictions at the earliest, many of the exit polls cannot consider samples from the second half of the last day of poll into consideration.

In the 2002 US midterm election filled with tight races, VNS pulled the plug on its exit polls after concluding that its computer analysis could not be trusted, which greatly slowed the television projections. Understandably, VNS was under intense pressure to avoid a repeat of the 2000 Florida debacle. But, the idea of exit polls was not abandoned in the US — they returned with a new polling service for the 2003 California gubernatorial recall election. “There is nothing unreliable about exit polls,” said Warren Mitofsky, whose firm collaborated with New Jersey-based Edison Media Research on that California exit poll.

Certainly, leading pollsters have an uphill task now to resurrect the lost glamour and credibility of exit polls. If not, exit polls might continue to look like mere skybus ads for the time being.

The writer is Professor of Statistics, Indian Statistical Institute, Kolkata