And I emphasize :
thoughts. So not for the kids or any poster who don't understand how polls work. So this thread is basically only for a handful of posters to post in.
- On the Dem side, the
entrance/exit poll (not sure how much of each) that's been published points towards a 49-46 Clinton advantage in the raw entrance votes, with O'Malley taking 2-3 and the rest uncommitted. I know the results being posted show a Clinton win with 49.8-49.6 against Sanders, but that's not how this process works. Polls were asking the electorate as a whole for their entrance choice, not the state delegate equivalents.
So when you take the RCP averages of Clinton 47.9, Sanders 43.9, O'Malley 4.3, take two points out of O'Malley, give one each to Clinton and Sanders, and you have your "real" entrance result. So the average of recent polls was pretty much spot on, as it very often is.
What's more, every poll counted in the last RCP average had the putative entrance result of 49-46 in their margin of error except, you got it, Gravis. DMR/Selzer and NBC/Marist both had the correct Clinton+3 margin, the former with a bit more undecideds. All things said, a very good night for the polling industry as a whole on the Dem side, once you know that you should consider all polls and at least average them, better still ponderate them, and not consider one lone poll just for what it says in its corner.
- On the Pub side, things look a bit more problematic. The very first teaching of this race is : you should poll as late as you possibly can if you want to detect final variations in a volatile race. Both polls that started polling on 1/29, however crappy the pollsters (Emerson and OpinionSavvy) may otherwise be, detected Trump's lead plummetting, and Rubio surging. On Rubio, only NBC/Marist had had him at 18% a few days before. Quinnipiac, by polling from 1/25 to 1/31, also got Rubio higher than the rest at 17%, I suspect from the data gathered on their two last days of field work.
Ironically, the rest of the candidates were pretty well polled by most firms, except the two late-comers : Emerson got Carson at 3 and Huckabee at 5, when OpinionSavvy got Paul at 9 and Bush at 5. So being late in the game can get you to assess the frontrunners better, but hurts you in the small figures, it seems. Which nobody cares about...
Finally, among the (somewhat) reputable pollsters, all had Cruz within their MoE, more or less. DMR/Selzer actually sort of had Trump in theirs too, though barely at 28%. NBC/Marist were close to having Rubio in theirs.
It was all a question of turnout scenario and voter screening. And momentum. When you look at the graph on the RCP average, and take into account that the middle point of all the field work for all the polls used in the last average is around 1/27 or 1/28, add 4 days and follow the curves and you know how it became a 3-way race.
- Now for NH : wait and see.
