Polls once again are not monitoring the public opinion, they are manipulating and socially engineering the opinion to agree with their expected poll results.
There is no way that a few thousand people can attempt to give the results of the entire 330 million Americans. The only way that poll results look accurate is because people conform to the poll results. Most of the people have no idea beyond the title of the poll and the results.
Take a very simple poll result
4 out of 5 doctors recommend brand X
So to the average person that means that 80% of All doctors recommend brand X. But they didn't ask 100% of the doctors they only asked a small sample size.
1. What do you know about the doctors in this poll. Nothing
2. You don't know what kind of doctors they are, for all you know they could be PHDs instead of MDs. You are inferring that they must be MDs.
3. You don't know how these doctors are familiar with Brand X, and whether they get a great deal from Brand X like free samples and other freebies.
4. You don't know where these doctors are.
5. You don't know the number of doctors polled.
6. You don't know how many doctors refused to take the poll.
7. You don't know the exact answer they gave to the ?, in fact you don't even know the exact ? asked, and in what context it was asked by the pollster.
8. You don't know why they recommend Brand X, or for what type of patients and conditions they recommend it.
9. You don't know what kind of results that Brand X had with their recommendations.
1o. You don't know the experience level of these doctors.
11. You don't know how brand x even fits into their patients and medical specialty.
12. You don't know whether Brand X is cheaper or more expensive than Brand Y. If X is cheaper maybe that is the reason for the recommendation.
There are so many things that you don't know about the poll. You infer a lot of it on your own. So vagueness and ambiguity are resolved by you most likely to ignore them, and just focus on the math results.
Remember statistics can be driven to any number of conclusions by the poll makers. And the statistical accuracy they are talking about is really the accuracy of their math, and not the accuracy of the answer to the questions.
Statistics are great for determining quality control in manufacturing processes because the processes are uniform. The quality is determined by the result of the process. So they can take a few samples and extrapolate the results across the process. But that is because the process and the material are uniform. People are not.