Giddy-up. Gotta keep that horse race going.
Prediction - Florida will be called early – for Clinton.
Something I’ve been wondering. Assuming they only poll “likely voters” and assuming that Dems do more early voting: as Election Day gets closer wouldn’t the potential universe of “likely voters” become more and more Republican-heavy? There are fewer and fewer Dem “likely voters” because more and more of them have already voted, making it inevitable that the polls would tighten.
I’m sure the people who do this for a living have factored this in somehow, Can someone explain how?
I’d like to know who answers a landline for a number they don’t know.
Yes, they do factor it in. They ask respondents if they’ve already voted. (And nothing is “likelier” than that.)
Selzer & Company, at A+, is the highest-rated of the recent FL polls included on 538. It shows Trump at +2.
Florida is really two states. The southern part is a state filled with refugees from the northeast. The northern part is filled with non-reconstructed sons and daughters of the old south. The state pretty much splits down the middle.
And Selzer says the reason it’s polling has been an outlier is because it is using a turnout model that is whiter and older than pretty much anyone else’s this year. Either she’ll be vindicated and we’ll all be sad, or she’ll have damaged her 24k reputation with some bad guesses a likely voter screen that filters solely on the basis of self-described voting intent, which other studies indicate is a pretty poor predictor of actual voting.. Pretty much all-in on it this year.
I will repeat that the greatly expanded use of LV screens and the apparently universal suppression of RV sample data is both inexplicable and introducing more, rather than less, uncertainty and it’s just bizarre given that the last couple of elections have demonstrated that LV screens make polling worse rather than better and that the questions traditionally used in them (self-declared likelihood of voting in particular) are sources of error rather than accuracy.
There is now such a huge divergence in the methodology used in internal and public polling that the two are categorically different. Which is why, for example, the Obama campaign knew with great certainty which states it was going to win and which it would lose with an astonishing degree of accuracy, in advance, even in states like NC that came down to less than 100K votes.
http://www.slate.com/articles/news_and_politics/victory_lab/2011/12/likely_voters_lie_why_private_campaign_polls_get_such_different_results_from_public_media_polls.html
Likely voter screens allow people like Setzer to put their thumbs on the truth. In the last week watch as she switches her model to something more conventional and comes up with something close to the final result.
PEC has FL at >80% probability of Clinton. So there’s that.
http://election.princeton.edu/electoral-college-map/
p.s. The chart is screwed up in my browser, with all the points on the left end and no horizontal axis labels or tick-points. Anyone else have this problem?
p.p.s. Hitting “zoom out” fixed it.
I don’t think she’s putting her thumb on the scale or doing anything conspiratorial. I have absolute faith in her integrity and to her commitment to respecting the integrity of her data.
What I do have a problem with–and I don’t recall her doing this before–is using a really stupid likely voter screen that consists entirely of asking people how certain they are to vote and dropping anyone whose answer isn’t “already voted” (okay, obviously, I have no problem with that one), “certain,” or “very likely.”
The 2008 Greenburg/Quinlan followup study of 12K+ previously polled people referenced in the Slate article I linked to indicated very clearly that this is a really terrible way of identifying likely voters. And yet, it seems to be the only screen she uses, apparently because she’s determined to prevent anything “unobjective” to influence her model.
Bottom line, if the Trump campaign and RNC and state Republican party voter suppression works, we’ll see Selzer’s turnout model. If it does not, we won’t. What we need is confirmation or not from other pollsters of any trend regarding Democratic turnout and likelihood to vote.
The reason for the comment wasn’t to forecast or promote any one outcome but to add information that TPM omitted, namely, what are the relative quality of these polls. I would have posted with the exact same template had the top-quality poll put Clinton ahead. BTW, 538 has an article on this that mentions Selzer previously did have Clinton wildly ahead – by as much as 18 points in their March nationwide poll – so it seems questionable to me to assume that their current outcome is only due to a Trump-favorable bias in their methodology.
Last night, MSNBC showed the projected results for an “if only millennials voted” election: Trump garnered a whopping 25 Electoral Votes, Clinton got all the rest. There may be hope for this country after all. (Not sure if there is hope for Florida, though.)