2016 Postmortem
Related: About this forumToo tight voter screens.
Yes, Hill is up in the polls but I wonder if her lead is even larger than the lead she has in the polls due to too tight voter screens. Gallup made that mistake in 012 and they were off by 5 points.
Nate Cohn is doing a North Carolina tracker and has discovered that 10% of respondents who said "it was not at all likely they would vote" actually voted.
Cicada
(4,533 posts)Some polls try to get at this by asking "what are the odds you will vote?"
They weight responses by that. TheRand Corporation poll does that and the LA Times tracking poll uses the Rand methodology.
I think a lot of the Trump surge in polls reflects his people moving up on the likely voter scale which count only those saying they are certain or very likely to vote. And on those polls they will miss the votes of the unlikely voters who favor Hillary.
DemocratSinceBirth
(99,710 posts)Don't let the L A TIMES folks mislead you. They aren't using Rand's methodology and are finding a different race:
Please Note:
Some people have confused the 2016 RAND PEPS and 2012 RAND Continuous Presidential Election Poll with the LA Times/USC Daybreak poll. The RAND and USC efforts are entirely separate and unaffiliated, relying on different panels and different weighting procedures.
RAND and USC both apply weights to ensure that results are representative of the overall population. However, RAND uses a two-step weighting approach for both the PEPS and the 2012 polls, while LA Times/USC combines those into a single step, splitting the demographic cells into smaller ones that can generate more extreme weights. Each incorporates self-reported 2012 voting behavior, but note that the PEPS results are not significantly different even without this adjustment, using only the standard demographic weights. Further, the RAND polls have both been conducted on the RAND American Life Panel, which has been collecting information on panel members for over a decade, while the LA Times/USC data uses a different panel established in 2014.
http://www.rand.org/labor/alp/2016-election-panel-survey.html
They have Clinton winning 44-35:
http://www.rand.org/pubs/research_reports/RR1792.html
Cicada
(4,533 posts)and that is pretty cool