Like double blind studies, polls should be a fair number of random samples of a those subjects being tested. But enough statistics studies and data are known that show that one cannot completely guard against “skewing” the results. Quantification problems aside, as well as “question item premise issues,” how does a pollster guarantee “un-bias?”  Here would be a help. Respondents who do not vote should be tossed or respondents should only be pulled from lists of voters. Also, respondents who pay no income taxes or have no payroll withholding, should be excluded. Sales taxes and DMV fees do not apply here. The reason being two-fold: One, if they are not voting they have pulled themselves from the process, and their opinions have no weight in an outcome. Two, if they are not filing or are not being taxed, they are not contributing to the various government coffers that effect the population.