It is interesting to see the MOE for each poll. For example, according to ARG, Fred is at 5% in Iowa (MOE 4) so be might be as high as 9%. Another poll (SurveyUSA) has him at 16% (MOE 4.5) so he could be 11.5%. This is not overlapping with ARG poll so clearly some poll - by definition - is incorrect.
I suppose it is very difficult to have a reliable polling in small states. Still, some trends can probably be spotted and average is typically quite good predictor. (”Very highly likely voters” has also been useful indicator - but not all polls use that).
Yeah, but that’s really stretching logic. Fred could easily be at near zero, and if you look at realclearpolitics and the TREND LINES for virtually every state, he’s lost between 1/3 and 2/3 of his support just in three weeks.
Polls in caucus States mean less than jack snot. Nobody but the campaigns, AFAIK, have access to the caucus members’ information, which is provided by each County caucus to the campaigns of their party. This info is released each week on Monday afternoons to the campaigns. It covers Sun-Mon daily updates of the prior week.
ALL of the polling organizations rely on telephone surveys of so called “likely” voters, not actual causcusers.
Sample verification testing, such as telephoning the same household on different days/times has proved telephone surveys are a matter of averaging out the five (for example) surveys each respondent has answered. Again, AFAIK, Rasmussen is the only one to use verification call backs.