Selection of sample is key to opinion poll accuracy: Psephologist

Why do opinion polls go wrong? While there could be many reasons, the fundamental one is that if you get the sampling wrong, the results go wayward.

Mathematician and psephologist Rajeeva Laxman Karandikar, who has been involved in pre- and post-poll surveys in about 40 elections in the last nine years, says the key to get your survey sample right is to ensure randomness. It can be mathematically proven that a truly random sample’s profile will match the population’s profile, notes Karandikar, a Director at Chennai Mathematical Institute and a winner of the Shanti Swaroop Bhatnagar award.

Karandikar feels that the sampling technique known as ‘circular systematic sampling’ is the best way to ensure randomness of the sample. In this method, you first pick up the clusters (in the case of Indian elections, States).

From the list of States you start picking up every, say, third state Starting from any number on the list. Then within each State, you choose every, say, fifth constituency from the list of constituencies in the State. Go on like this until you arrive at a list of voters.

That was the method that the Delhi-based Centre for Studies in Developing Societies followed. The news channel CNN-IBN has hired the Centre for doing opinion polls. Karandikar analyses the data.

So, what about this time?

In a chat with Business Line , Karandikar said there are other factors that could cause pre-poll survey results go awry. For instance, a survey gets the mood of the entire population, while on the voting day, some sections do not vote at all, skewing the results. Of course, some events sway voters’ sentiments.

“When parties hop alliances, leaders hop parties, why should the supporters be loyal?” asked Karandikar.

This time, however, he sees little reason to believe that the elections results will be different from the opinion poll indications. Turnout at the booths has, by and large, been high – which means one reason for survey waywardness is not there.

“Traditionally, high voter turnout has favoured the BJP,” notes Karandikar. It is typically the BJP supporter – the urban middle-class voter – has baulked at standing in a queue in the sun. Also, there has been no event so far that could cause voters change their minds, he points out.

“There is a Modi wave,” Karandikar says, noting that a smaller pro-Modi wave has got lifted up by a strong anti-Congress wave. People in rural Tamil Nadu have picked Modi for the Prime Minister in the CSDS surveys, he said.

Karandikar favours post-poll surveys – not exit polls, where it is impossible to be truly random in picking up samples. CSDS has been doing post-poll surveys, going door-to-door in the constituencies after voting. The results would be out after all the polling is over but before the counting begins, he said.

Post-poll surveys are remarkably accurate, Karandikar says. For instance, CSDS post-poll survey showed a massive victory for the Congress in the 2011 Assembly elections in Assam, at a time when nobody, including the Congressmen, believed that could happen. Congress won 78 of the 126 seats; favourites Asom Gana Parishad got 10.

COMMENT NOW