1. Incentive driven Opinion
The most popular opinion poll today is being published by Times Now and conducted by C Voter. While they are actively talking about the number of seats a party is going to win in a particular state, they are not talking about the possible error margins. After all they are trying to predict how the nation will vote and not just that they are talking about the seats in a constituency and share of votes on the basis of views expressed by an incentivized sample. The incentive driven nature of the sample brings in huge bias especially when you are trying to get the opinion of a nation.2. Almost Prophetic
As long as the opinion poll sticks to its premise i.e. to indicate an opinion, it is good but currently the manner in which they are being presented is nothing short of a foretellers' prophecy. While it is OK to seek perceptions and opinions of people, if you start predicting the number of seats a political party is going to win in an election, there is a problem. Now you may argue that the problem will be for the party lagging in the opinion polls, which is a fair.3. What about the Voter turn-out?
A big factor in election results is the voter turn-out. The predictions based on opinion polls discount the impact of voter turn-out, which again induces a huge error margin. Different sects in a constituency have differing opinions and differing turn out ratios. Thus it is improper to make estimations based on just opinions.4. The Undecided voter
Another reason I doubt the scientific validity of these polls is that they have big month on month changes in their projections. In a way it means that India has a high uncertain voter base. Is it so? It could also be because the sample, which is being used for estimation has a high percentage of undecided voters. It also assumes that this voter base will go out and vote.5. The loyalist
Does the sample used for predicting seats won by a party also have a good representation of staunch loyalists of a particular party? How many RSS workers, rickshaw pullers, laborers etc. are there in the sample. It could only be the educated lot is part of the sample, which is biasing the outcome. This is again not the case.Opinion polls in its current state are extending beyond their boundaries into a territory where their accuracy is questionable and their predictions misleading. As long as they indicate opinion, they are alright but when you start predicting the number of seats then there accuracy and process becomes highly questionable. The impact gets magnified when leading news channels start actively advocating such polls and start questioning party spokesmen on the basis of these projections. Opinion polls should perhaps stick to to giving opinions and not predicting the election outcome.