Straw polls and crystal balls

Just about two weeks ago, CNN-IBN hosted a TV show analyzing the results of an opinion poll conducted by the venerated CSDS. A number of politicians and subject matter experts including Yogendra Yadav, CSDS supremo, were on the show.

The very first thing that struck us were the outfits of the spokesmen of India's two biggest political parties. BJP spokesman Ravi Shankar Prasad wore a dark brick red Kurta over white pajamas. A garish dress no doubt for such a formal occasion. But look at what Congress spokesman Abhishek Manu Singhvi is wearing! A black leather short-sleeved shirt, with yellow-gold patterns running along the buttons, around the collars and across the front pocket. The last time we saw someone in an outfit like that was Bappi Lahiri.

Anyway.

Apparently this was the biggest opinion poll of its kind, a total of 18,000 respondents were surveyed all across the country. In the words of Yogendra Yadav (note: all transcripts ours):

Yogendra Yadav: This being a very crucial occasion and mid-term elections increasingly almost certain , we wanted to make sure that we wanted to do a poll which is almost a mock mid-term poll ..... here it is then we spread it out this is truly representative..... rural urban the same proportion as in country. 73% rural, 11 - 12% Muslim, 18% Dalits, 45% Women - this is microcosm of India.

The opinion poll revealed a vote difference of almost 10% between UPA and NDA. This translated into an immense difference in terms of seats (267 UPA, 133 NDA). Host Rajdeep Sardesai asked Prasad to react to this.

Ravi Shankar Prasad: Rajdeep, we have also learned a little experience of watching your surveys. In 2004 these projections gave us a big lead and there was no hostility against the NDA Govt. And what the results came about? I very acutely recollect one comment from Yogendra that a party in power initially gets a big boost in these surveys but when the concrete election campaign starts ...

Rajdeep then asked Yogendra Yadav to comment on this.

Yogendra Yadav: Absolutely and it's a fair point and this particularly operates in those states where UPA is in power at the center and also at the state level so there may be built an element of bias, so we have tried to correct it, we have carefully tried to eliminate but we can not give guarantee that this will not happen..

Fast forward several minutes and now we come to the point where they are discussing the predictions for Gujarat assembly polls.

Rajdeep Sardesai: One man, and this is another significant finding of our poll, seems to be bucking the trend of anti-incumbency. Remember he faces dissidence within his party almost since the time he came to power... severely criticized by many political analysts for his roll during Gujarat 2002, now as Gujarat prepares to go to the polls look at what our figures are saying.... Narendra Modi is holding firm, he is likely to get we believe 54% votes as against Congress' 43% if elections in Gujarat assembly are held tomorrow. Yogendra, any riders here?

Yogendra Yadav: One rider is one given already by Mr. Ravi Shankar Prasad very important one, that in any pre-election poll, the ruling party, there is a bias in favor of the ruling party which we need to be careful about and second we must not underestimate [] what is happning to BJP at the national level .... its inability to form an alternative in presence of problems in the Congress, is happening to Congress in Gujarat.

This got us entirely confused. What is it Mr. Yogendra Yadav, was a correction really done to prevent a known bias or was it not?

What a difference a mere two weeks makes. Perhaps these results do not appear so valid in the backdrop of the Ram Setu controversy. During the same time period, C-Voter conducted another opinion poll and arrived at different results. For comparison, also check out qualitative predictions of an independent blogger.