Polls are inherently flawed instruments. While they provide endless fodder for discussion, the flaws make them the proverbial 'grain of salt', as in take this with...
Statisticians can knock themselves silly constructing a fair sample, and can 'prove' that it's representative of the population on the whole, but it's not and never will be. Pollsters 'assume' the sample population they have selected wants to answer their questions, and answer them honestly.
First, many people simply hang up the phone if a pollster calls, or walk away if a pollster approaches them on the street. Thus, at best pollsters get a statistically significant sample of people who are willing to talk to them. That leaves out lots of people, maybe more than half. Still, pollsters will project the answers on to the overall population, but it hardly can be claimed to be representative of the people who purposely avoided them.
Second, not all people will answer honestly, and that for numerous reasons. Maybe the pollee assumes what the pollster wants to hear and gives them that answer. Maybe a wag gives them the opposite: assumes what the pollster wants to hear then tells them the opposite. Then there are people who prefer not to reveal their true belief, but are still willing to listen to a pollster, perhaps out of curiosity or politeness.
So pollsters are left with a limited and faulty true sample of a society and may be getting false or deceptive answers. The only way a poll could be considered accurate is if EVERYONE was willing to answer, and answer honestly. Then the sampling method would make sense and be truly representative.
Polling isn't going to go away, because there's money in it and most of us are amused by polls, but since there really isn't any way to correct for the issues noted here, the true 'margin of error' is likely to be a multiple of what is given.
Thailand is the 54th happiest country in world
in Thailand News - Discussion
1.Lie. 2.Big fat lie. 3.Statistics.