The business of opinion polls / pre poll projections is big in India now. This is very much in demand in Lok Sabha and State Assembly Elections. A lot has been written on whether it should be banned as it impacts voting pattern of people. It shapes public opinions and helps political parties gauge the mood of the people. But are they accurate? In the latest 4 state assembly elections, for 3 states every opinion poll was correct. But these states were done deals. No survey was needed to say BJP was winning in Madhya Pradesh, Chhatisgarh and Rajasthan. Delhi was tricky, very close and almost all opinion polls got it wrong barring one.
So it brings us back to the moot question are these opinion polls correct? For the last three elections – 1999, 2004 and 2009, opinion polls have gone horribly wrong. Especially 2004 was a blot for the opinion poll industry. Everybody predicted for BJP win with comfortable majority. BJP lost badly and got half of the seats. What is wrong with these opinion polls?
1. The sample size is too small.
In this elections sample sizes for most polls have been 50k. NDTV claims a sample size of >2 lakhs. This is <0.1% of the total voter base of 80 crores. Even if you assume 50% voting, this is still <0.1%. How can such a small sample predict the voting trends and patterns?
2. All constituencies are not covered.
India has a total of 543 constituencies. Opinion polls cover 250-300. NDTV the biggest opinion poll (as they claim) covered only 350 constituencies (<65%). Without covering 200+ constituencies, how can voting pattern of those seats be predicted?
3. Opinion polls assume people would answer survey questions honestly.
This is the premise of the whole exercise. In India people hesitate to discuss openly about their political preferences. Plus some people deliberately give wrong answers for the sake of fun.
4. Poll results can be manipulated. Survey questionnaires can be tweaked to get the result you want.
Plus lately there have been various allegations that most of these polls are paid for. Survey / Research companies can be sponsored to carry out polls by political parties. A new type of polls have emerged. That by political parties. Like the one carried out and advertised by Aam Aadmi Party in Delhi state elections. These are definitely manipulated.
5. Proportion of rural population in sample size is small:
Majority of people who actually go out and vote are from rural areas. The situation there could be entirely different than urban areas. Because of sheer costs involved, I doubt whether surveyors actually are able to reach out to rural people. Majority of urban people don’t vote so their opinion doesn’t matter much.
Pollsters extrapolate the results of the 250-300 constituencies they cover and come up with the whole country picture. It is more of an intelligent guessing based on trends and current mood. A lot of readers ask whether projections carried out by Politicalbaaba are backed by surveys. Answer is no. I have a network of people / followers from whom I take feedback and give my projections. It is in no way inferior to any opinion poll carried out by channels / newspapers etc.
Poor Track Record of Opinion Polls
2004: Everybody predicted a NDA win. NDTV / Indian Express gave 287-300 seats, India Toady 330-340 seats. Hindu in an article before elections (19th April, 2004) calculated an average of the polls as 271 seats for NDA. Actual result 181, 90 seats less than projected. Congress was projected to get on an average 150 seats, they got 218.
2009: Most polls expected Congress led UPA to emerge as the largest combination, but the margin between UPA and NDA was projected to be 20-30 seats. UPA 200, NDA 170-180. Actual result UPA 262, NDA 159, margin of >100 seats.
I couldn’t get 1999 Lok Sabha opinion polls on net. The industry was not very big then. However as per a blog by Prashant Panday (http://blogs.timesofindia.indiatimes.com/the-real-truth/entry/last-3-poll-forecasts-all-wrong-india-today-now-forecasts-an-nda-win), India Today predicted in Dec. 1998 that the BJP+ would get 135 seats while the Congress+ would get a huge 305. Exactly opposite of this happened. Couldn’t verify the authenticity though, presented for illustrative purposes.
2014 Opinion Polls
In 2014, opinion polls predict a NDA led victory. BJP led NDA is expected to get 220-260 seats as per most surveys. Will history repeat itself again? Only time will tell…. But one thing is for sure, opinion polls are not any way better than anybody’s (your or mine) guess. Politicalbaaba predicts 230 seats for NDA (excluding TDP).