2016 Postmortem
Related: About this forumDo public polls even have a 'turnout model?'
The spin from Money Boo Boo's camp is that Obama and the public pollsters have an erroneous turnout model.
But, don't public polls simply call people according to their census stats ask them if they are registered and likely to vote?
In other words, they don't have a turn out model, their projected electorate is simply what respondents tell them.
Or am I missing something?
DemocratSinceBirth
(99,711 posts)Because their robocalls generate reams of raw data that then needs to be refined.
My knowledge of statistics is cursory. But if you got to votamatic or princeton consortium the hosts might answer your question.
Tutonic
(2,522 posts)Ann's crying like a baby and Eddie is hiding out. Whatever his handlers (men in white coats) announce at this point you should believe the opposite. The man has a fool for a client--himself. He and Karl Rove are two of the biggest losers to lead a national party. You hear from Priebus lately? Even he can smell the stench on Money. Chris Christie--found a way to distance himself--he could smell that stench standing on the Jersey shore. By next weekend, Money will be reduced to a final skit on Saturday Night.
PoliticAverse
(26,366 posts)If you look at the explanation they will usually tell you what method they used, one based on just
registrations or one that tries to guesstimate who will actually vote.
Here's a story that discusses the 'likely voter' issue and the methods pollsters used to determine likelihood:
http://www.huffingtonpost.com/2010/10/05/likely-voters-how-pollsters-choose-them_n_751560.html
smorkingapple
(827 posts)Despite all evidence to the contrary.
DemocratSinceBirth
(99,711 posts)They should put down the Cheetos, get out of their mom's basement, drive, bicycle, or walk to their nearest state/community college and take STA 101.
Cicada
(4,533 posts)I guess most polls assume those those who score as "likely voters" will turnout. But their calculation of "likely voter" may depend on their expected percentage who will vote.In a nonpresidential year you probably have to score real real high on their likely voter questions to be included, but you can be "likely" with a lower score in a presidential year. Gallup has 9 likely voter questions. In a nonpresidential year they may score as likely only those who get 8 or more "correct" answers, versus maybe seven "correct" in a presidential election year. Also, internal polls probably have more categories than "likely" who they assume will all vote vs not "likely" who they assume will not vote. In fact a certain percentage of likelies will not vote and some not likelies will vote. So a sophisticated poll will take a different percentage of each score on a likely voter scale as likely to vote. Pew once compared what respondents said about whether they would vote with actual voting records (they got names and checked voting records after the election) and found MOST who said they would not vote actually did vote. So the campaigns know stuff like that, probably based on their own prior research, and DO have turnout models, beyond just measuring likely voters vs nonlikely voters.
geek tragedy
(68,868 posts)the campaigns would have more of a true model?
I'm very skeptical that Republican pollsters understand Latino turnout dynamics better than Obama or Latino Decisions.