A closer examination shows that the surveys conducted by Asianet-C Fore and CNN-IBN-The Week were post poll surveys and not much are known for the methodology of all the others except Star News. Star News election survey was exit poll.
The difference between them is that, in the case of the former, the survey is conducted after a few days of the actual poll, where as in the latter, the survey is conducted on the poll day itself right from the polling station itself. This aspect has serious implications for the predictions. In the case of post poll surveys, the survey agency can take random sample in the most theoretical manner, that too in the most religious manner, since they are selecting the respondents from the electoral roll itself and tried to contact them.
But in the case of exit polls this is not possible. The survey agency can select respondents, however random it may be, only from the turnout of voters to the polling station. Even though it is assumed that the survey agency did not take respondents selectively, still they took them from the polling station itself. It clearly violates the principle of random sampling in the purest form. There can be a pattern for the voter’s turnout at the polling station.
For example, when voters come, they come in groups, sometimes on the basis of family, friendships, political affiliations etc. There will be high chance of homogeneity for the voter turnout at any point of time. Any exit poll survey conducted from such a voter turnout would highly deviate from the principles of random survey.
But in the case of post poll survey, if the agency wants a pure random sample, they could do that. Hence, the predictions of them would be more reliable. In the case of CNN-IBN-The Week survey, as pointed out by Prof. Yogendra Yadav, the methodology was the most scientific one which religiously adhered to the principles of random sampling. Hence the predictions of it possess more accuracy. The methodology of Asianet C-Fore survey was not explained as what was done by Prof. Yogendra Yadav. Not much can be said with respect to that, hence.
But one thing stands out. Both predict a tough fight between the LDF and UDF. Comparatively all the other surveys predicted big wins for the UDF including Star News survey. The reason for big win prediction by Star News survey is under suspicion for the flaws in the methodology of exit poll.
Both the post poll surveys (Asianet-C Fore and CNN-IBN-The Week) found out the favourable effect of VS Achudanandan factor in the poll. The reason is that he was the most popular chief minister candidate in the surveys. That is the reason for predicting a turnaround from the overwhelming trend of rule change witnessed for the last several decades and gave a close fight. If that is the case then that trend would give LDF a majority.
The reason is that, when there is a trend, especially a one that emerged in the late periods of the campaign, it cannot be fully captured by the surveys. People would be in confusion as to know which way the overall trend moves. A significant number of the voters may not divulge the details or may not give correct responses in surveys. In the CNN-IBN-The Week survey it is worth noting the fact around 11% of the respondents did not reveal details. This fact corroborates the above mentioned conjecture.
What boils down is that the latent trend, if such one was there, would give the necessary votes to the LDF for getting the simple majority. Otherwise there may not have a close fight and UDF would get a big victory.
