Making sense of surveys

I was looking for numbers to support/refute my simple, not so original hypothesis about Kerala's political dynamics. I tend to mostly believe CSDS data: because I mostly find Yogendra Yadav's analysis agreeable. At any rate, he is the better than the clueless TV journalists and newspaper editors like Shekhar Gupta and Veer Sanghvi.

However, in the data that appeared in yesterday's Hindu: it says the LDF lost 13% of Nair vote between 2004 and 2006, and the UDF gained only 4% of the Nair vote. So the 9% would have gone to the BJP or independents. The BJP's total votes came down from 10% to 6%, and Nairs and upper caste Hindus are the social groups they find some acceptance, thus "ungaining" say: 10% of total Nair vote. Now we have about 20% of Nair votes "swinging", yet the UDF has gained only 4% (as per CSDS).

The overall polling percentage increased by about 2%, so nothing there to account for a 10% of a group that makes 13% of total electorate not voting at all . 10% of 13% of 73% of 2.2 crores is about 2 lakhs - the most famous "independent" Neelalohitadasan Nadar - who might have mostly got Nadar votes - got only 1 lakh votes(only 50000 Nairs would have voted in TVM anyway, and mind you, Shashi Tharoor is a candidate that would have absorbed all upper caste votes, and also, we have a stronger BJP presence in TVM). Considering the BJP's swings as well, we have to explain for at least 4 lakh Nair votes.

Apart from Nairs, the only other group that shows a major swing as per the survey is the Christian group (13%). The rest, with swings of 5% or less, and considering the sampling errors of the survey, throw very little in terms of "which group voted for whom differently this time". I don't know what to make out of the survey, and to what extend this kind of data can be used to gain any insight into what actually happened.