Reply #4: Every poll makes assumptions.

[Igel]

Every poll has to.

If you take a poll at face value without looking at the assumptions that go into the model and how they use their model to adjust the data they get from interviewing people, you're making a mistake. You have no idea what the number means, no basis for evaluating the information you're given.

A lot of people like to obsesses over margins of error. These are calculated errors and assume that the model is correct but that the sampling may be non-random. Errors in assumptions don't show up in the margins of error. They're not quantifiable. They're ultimately "unknown unknowns"--either as to quality or quantity.