Why opinion polls are bad for elections

February 6, 2015

(Any opinions expressed here are those of the author and not necessarily those of Thomson Reuters)

Opinion polls are released by the media before every election in India, where they predict the winner days before official results are declared. The assembly election in Delhi, which has witnessed one of the toughest electoral battles in the country, has also been subjected to a barrage of media polls which are split in predicting the winner.

While six opinion polls declared that the BJP will win a majority of the 70 seats, four predict an AAP victory while one said it could go either way. This once again highlights the issue of election surveys being reduced to a media gimmick, where political parties and media and business houses with vested interests try to influence voters through opinion polls.

The media, which regards itself as a modern day Nostradamus, confuses voters by forecasting the results and this is an impediment to conducting a free and fair election. Even Prime Minister Narendra Modi trashed opinion polls as “lies”, pointing out that the AAP had claimed they will win over 50 seats in the last assembly elections in 2014, but could not even manage to get a majority.

It becomes pertinent in this context to trace the historical roots of election surveys and why they fail to fathom the correct mood of the electorate.

Delhi Assembly Elections 2015 – Overview of Seat Predictions:

Seat Forecast BJP AAP Congress
HT-Cfore 31-36 31-36 2-7
India TV-C Voter 36 31 2
ABP-AC Nielson 29 35 6
NDTV 29 37 4
IBN 7-Data Mineria 36 27 7
India Today-CICERO 19-25 38-46 3-7
ET-TNS 28-32 36-40 0-3
Zee-TRF 32-36 30-34 4
The Week-IMRB 36 29 4
News Nation 31-35 30-34 3-5
5Forty3 40 21 8

Note: The seat predictions are based on pre-poll surveys.

A quick recall of opinion polls in India reveals that the media’s obsession with election forecasting started on a massive scale two decades back. Seat predictions of past Lok Sabha elections show that opinion polls got it right in 1996, 1998 and 1999, which provided a big boost to the polling industry.

Seat prediction was then heralded as a scientific method of forecasting the electoral fortunes of political parties. However, the success was short-lived as poll predictions in the 2004 Lok Sabha elections were completely wrong and the credibility of the polling industry took a nosedive. The situation went from bad to worse in 2009, where pollsters failed to capture the mood of voters and the Congress-led UPA formed the government for a second consecutive term.

The aim of an election survey has been to study voting behaviour and attitudes of the electorate and provide trends during elections. But the basics of election studies have been pushed aside and opinion polls and seat predictions have become sinisterly synonymous with each other.

The polling industry got an opportunity to re-establish its lost credibility during last year’s general election, and a flurry of exit polls predicted that the BJP-led NDA was set to cross the majority mark. The tone and tenor of the seat forecasting by poll pundits and media houses appeared so definitive that the actual announcement on May 16 seemed only a formality. However the results showed that pollsters once again floundered badly as none of them (bar one) predicted that the BJP would get a clear majority on its own (282 out of 543 seats).

General Elections 2014 – Overview of Seat Predictions

Seat Forecast BJP allies Congress allies Others
News 24-Chanakya 340 70 133
India TV-C voter 289 101 153
ABP-AC Nielson 281 97 161
NDTV-Hansa 275 111
CNN-IBN 270 – 282 92-102
India Today-CICERO 261-283 110-120 150-162
Times Now-ORG Marg 249 148 146
Actual Result 336 58 149

Note: The seat predictions are based on exit poll except for CNN-IBN, which used a post poll survey.

The stakes are high for the polling industry and the following factors can play a detrimental role in the accuracy of their predictions for the Delhi election:

1. The high-voltage communication strategy of the BJP and AAP has created a high level of awareness among the voters, and polls run the risk of recall bias. The presidential-style campaigning by Arvind Kejriwal and Kiran Bedi seems to have mobilized a large number of voters but whether it gets translated into votes is a tricky factor to capture in opinion polls.

2. Delhi is witnessing a triangular competition and different political parties dominate different areas of the city. It is quite a challenge to convert the vote share into actual seats. A review of seat prediction in states having multi-polar contests reveals that predictions tend to go wrong more often than right.

3. The Congress party, which bore the brunt of anti-incumbency and the anti-corruption movement, may regain some of its lost votes and dent the fortunes of either the BJP or the AAP.

4. Conducting opinion polls in metropolitan cities like Delhi has inherent risks as the non-response error is very high and surveyors may resort to fake entries and data fudging to complete their targets and deadlines. This may lead to serious error in computing the right vote shares of political parties.

5. Lastly, opinion polls can be right in capturing the correct vote share of various political parties in the elections, but the prediction models can have in-built errors. Likewise, the forecasting methods could be right but the exit polls can be flawed and have various kinds of survey error.

The call for a blanket ban on opinion polls during elections will grow much louder if the forecasts for Delhi turn out wrong. However, opinion polls should not be banned as they are done by the media worldwide to gauge the popular mood of voters. What needs to be banned by the election commission is seat forecasting before the elections. It is time to lift the mystery over the seat prediction models used by polling agencies and bring them into the public domain. The current trend of using seat predictions as a tool of communication to create a bandwagon affect for a particular party should be stopped immediately.

No comments so far

We welcome comments that advance the story through relevant opinion, anecdotes, links and data. If you see a comment that you believe is irrelevant or inappropriate, you can flag it to our editors by using the report abuse links. Views expressed in the comments do not represent those of Reuters. For more information on our comment policy, see http://blogs.reuters.com/fulldisclosure/2010/09/27/toward-a-more-thoughtful-conversation-on-stories/