Opinion polls, though popular, are still not reliable in India
Almost all poll pundits say it is highly unlikely that opinion polls can influence the outcome of elections. The only way surveys can influence the voter is if all of them predict the same result. The fence sitters are then likely to turn to the winning side.
In mid-August, the C-Voter - India Today poll predicted that Narendra Modi remains the people's first choice for the post of Prime Minister but he has yet to convert his popularity into votes for the Bharatiya Janata Party (BJP). C-Voter estimated a Lok Sabha tally of 155 for the BJP-led National Democratic Alliance (NDA) in the 2014 elections, about 18 seats more than the ruling United Progressive Alliance (UPA). A few days after the survey was made public, Yashwant Deshmukh, the founder of the polling agency, C-Voter, received a rather unusual request.
Deshmukh was requested by ICICI Bank to visit Mumbai for two days and meet 18 chief investment officers of foreign institutional investors. Global rating agency Standard & Poor's also sought an interaction with Deshmukh and almost a fortnight later decided to hold its sovereign rating of India until after the Lok Sabha polls next year.
Deshmukh held separate two-hour-long meetings with the CIOs and the S&P representatives to discuss his survey findings. Now, he says he has been invited to address another such investor meet in Mumbai in December. This was the first time the investor community has sought a pollster to discuss the various post-election scenarios. In the past, such meets have been largely addressed by news editors.
The above example highlights the growing clout of pollsters in India and reflects the increasing demand for and value of political surveys - from global agencies like ACNielsen to local pollsters like C-Voter, they are all taking a shot at predicting who will form the next government in the largest democracy of the world. Surveys are getting commissioned by media houses, political parties and even candidates. Deshmukh got two calls from a leading television channel for a Chhattisgarh state survey during his conversation with Business Today. This is the kind of frenzy over surveys, whose reliability and accuracy is questioned by many.
The ruling Congress is now the latest to join the chorus against opinion polls. The main opposition party BJP is quick to portray the Congress stand as a case of sour grapes - most polls show the party losing ground across the country. Meanwhile, the Election Commission has already sent a proposal to the law ministry to ban pre-poll surveys.
There has been a proliferation of home-grown pollsters in the last few years, with smaller agencies such as Cicero battling it out with bigger ones such as C-Voter and The Centre for the Study of Developing Societies (CSDS). While increased mobile and net connectivity has made it easier for agencies to conduct all kinds of surveys - on phone and online - there is no consistency about methodology. Most of these polls are blamed for being largely urban and male-centric. "You need to reach out to remote locations, which add to the costs. So, people try and take shortcuts," says Sanjay Kumar, a professor at CSDS. There is a lot of online data collection, which is very unscientific, according to V. Surya, Vice President, IMRB International. "Sample sizes are hardly anything. There is a question mark on reliability. Many of the surveys are not talking about how they collected their samples," adds Surya.
Indeed, in the last two Lok Sabha elections almost all pollsters got it horribly wrong. In the 2004 elections, most polls predicted the NDA would come to power and they got it wrong. In the 2009 elections, most polls predicted the winning alliance correctly, but were off the mark on seat projections. "Polls were saying that the UPA would reach 180 to 185, nobody expected it to cross 200," says CSDS's Kumar. The final tally of the UPA was 262 seats.
HOW IT WORKS
In times of competitive bidding for projects, pollsters are charging as low as Rs 10 per question to a respondent. Survey costs can range between Rs 75 and Rs 100 per respondent or even more. There are various variables that determine the accuracy of a survey - timing, sample size, the type and order of questions, and training of the interviewer. A large complex survey can require close to three hours per respondent. "Asking the right questions is an art. Intensive training of the interviewer, body language makes a difference to the responses," says Pronab Sen, former chief statistician of India. Most polling agencies do not share their raw data, which makes it almost impossible to cross-validate the findings. "Data is a black box which only the agency knows," says Sen.
Now, the big question - can opinion polls influence the outcome of elections? Highly unlikely, say almost all poll pundits. The only way surveys can influence the voter is if all of them predict the same result. The fence sitters are then likely to turn to the winning side.
With a fledgling ecosystem, polling agencies have to rely on media and even political funding to really thrive. Indeed, some pollsters in India now have a political connect - Deshmukh is the son of late Rashtriya Swayamsevak Sangh leader Nanaji Deshmukh while Yogendra Yadav, professor at CSDS, is now a member of the Aam Aadmi Party. "I want to remind people who question my credibility that I was the one to predict a Congress victory in Karnataka," says Deshmukh. In a country of India's size and population, then, it's difficult to say which way the political wind will blow until the day of voting. In the absence of an established scientific methodology followed by most pollsters, the surveys can at best be a check on the mood of the nation.
(Follow the author on Twitter: @shwetapunj)
Deshmukh was requested by ICICI Bank to visit Mumbai for two days and meet 18 chief investment officers of foreign institutional investors. Global rating agency Standard & Poor's also sought an interaction with Deshmukh and almost a fortnight later decided to hold its sovereign rating of India until after the Lok Sabha polls next year.
Deshmukh held separate two-hour-long meetings with the CIOs and the S&P representatives to discuss his survey findings. Now, he says he has been invited to address another such investor meet in Mumbai in December. This was the first time the investor community has sought a pollster to discuss the various post-election scenarios. In the past, such meets have been largely addressed by news editors.
You need to reach out to remote locations, which add to the costs. So, people try and take shortcuts
Sanjay Kumar
Professor, CSDS
The ruling Congress is now the latest to join the chorus against opinion polls. The main opposition party BJP is quick to portray the Congress stand as a case of sour grapes - most polls show the party losing ground across the country. Meanwhile, the Election Commission has already sent a proposal to the law ministry to ban pre-poll surveys.
There has been a proliferation of home-grown pollsters in the last few years, with smaller agencies such as Cicero battling it out with bigger ones such as C-Voter and The Centre for the Study of Developing Societies (CSDS). While increased mobile and net connectivity has made it easier for agencies to conduct all kinds of surveys - on phone and online - there is no consistency about methodology. Most of these polls are blamed for being largely urban and male-centric. "You need to reach out to remote locations, which add to the costs. So, people try and take shortcuts," says Sanjay Kumar, a professor at CSDS. There is a lot of online data collection, which is very unscientific, according to V. Surya, Vice President, IMRB International. "Sample sizes are hardly anything. There is a question mark on reliability. Many of the surveys are not talking about how they collected their samples," adds Surya.
Indeed, in the last two Lok Sabha elections almost all pollsters got it horribly wrong. In the 2004 elections, most polls predicted the NDA would come to power and they got it wrong. In the 2009 elections, most polls predicted the winning alliance correctly, but were off the mark on seat projections. "Polls were saying that the UPA would reach 180 to 185, nobody expected it to cross 200," says CSDS's Kumar. The final tally of the UPA was 262 seats.
HOW IT WORKS
In times of competitive bidding for projects, pollsters are charging as low as Rs 10 per question to a respondent. Survey costs can range between Rs 75 and Rs 100 per respondent or even more. There are various variables that determine the accuracy of a survey - timing, sample size, the type and order of questions, and training of the interviewer. A large complex survey can require close to three hours per respondent. "Asking the right questions is an art. Intensive training of the interviewer, body language makes a difference to the responses," says Pronab Sen, former chief statistician of India. Most polling agencies do not share their raw data, which makes it almost impossible to cross-validate the findings. "Data is a black box which only the agency knows," says Sen.
Sample sizes are hardly anything. Many of the surveys are not talking about how they collected their samples
V. Surya
Vice President, IMRB International
With a fledgling ecosystem, polling agencies have to rely on media and even political funding to really thrive. Indeed, some pollsters in India now have a political connect - Deshmukh is the son of late Rashtriya Swayamsevak Sangh leader Nanaji Deshmukh while Yogendra Yadav, professor at CSDS, is now a member of the Aam Aadmi Party. "I want to remind people who question my credibility that I was the one to predict a Congress victory in Karnataka," says Deshmukh. In a country of India's size and population, then, it's difficult to say which way the political wind will blow until the day of voting. In the absence of an established scientific methodology followed by most pollsters, the surveys can at best be a check on the mood of the nation.
(Follow the author on Twitter: @shwetapunj)
