As I see it, the problem with polls is that they rely on a number of factors that are not givens, and that the results are accepted and treated as if they were a truly representative result, when such is not necessarily the case.
One of the basic problems is that people are NOT a homogenous mixture. We are individuals, who may or may not be found in groups of people with similar opinions and experiences.
Think of it as a chemical make up tank without being stirred. All the different chemicals are dumped in, but they are not stirred and properly mixed to a uniform content.
Sample size and LOCATION make a difference in the result. Sample 1oz in one spot and you get one result. Sample 4oz in that same spot and you MAY get a different result. Sample in a different spot and you may get yet anther and different result.
Sample 1,000 or 50,000 in the urban metro areas and I'm willing to bet your results will be different than sampling the same number of people in rural areas.
Another thing about surveys, which is not found in chemical sampling, chemicals don't lie to throw off your survey.
People do.