As we all know this is not a typical election cycle.
Polls are good but as the season progresses they more and more become likely voter polls and those are only as good as their likely voter screens are at accurately appraising who is and who is not a likely voter in a particular cycle. If a demographic behaves differently than they have in the past then the screen may be off. And while likely voter status is partly ascertained simply by asking whether or not someone is intending to vote and how sure they are of that, it is also usually based on past voting behavior and on demographics.
So, IF, for example, somehow Trump does get many more eligible voters to the booths who have not bothered to vote in the last few cycles, without losing other groups too much, then he may outperform his polling. Likewise polls can be off because they misestimate turnout of the Hispanic population up or down, or assume turnout of previous White college educated voters that are Trump leaning that fail to show.
Who knows? Maybe more than one factor.
But those factors will play out differently in different states with some states’ aggregated polls closely calling the results and some guessing wrong on who actually comes out to vote this time.
Any guess on which state will miss by the most (even if the result of win/loss for the EVs is the same)?
I’m guessing Utah. I don’t think previous voter behavior is going to be a good predictor of who votes this time there.
Which states win your best guess to have their polls be least accurate?
To make sure the question is clear - not asking which current polls will be most off but which final rolling aggregates will be. For simplicity we can use the final RCP rolling averages. Of course only the states that had polling need apply.
For reference the biggest misses in 2008 by that metric apparently were NM by 7.4 (Obama won by 14.7 instead of the RCP rolling avg of 7.3); IA (Obama lost by 9 instead of 15); and NV (Obama won by 12.4 instead of 6.5). Alternatively in 2012 NM polling was right on the money. None altered which way the state went. (OTOH NC was within 0.8 but flipped as a result, RCP rolling average was McCain +0.4 and final was Obama 0.4.) Most states in 2008 were within 2 or 3 of their RCP rolling average, and likely 538’s approach did better.
Anyone want to add in their guess for how far off the final national polling predictions will be feel free. I am guessing they will underestimate Clinton’s margin by a few points.
I’m guessing that Georgia will be more pro-Trump than the polls will look and some Democrats will get themselves excited unnecessarily. I can’t imagine a lot of places have a lot of experience polling Georgia since there’s rarely anything competitive going on there and so I could see some learning-curve swings that average to more competitive than it really is.
Of course, I’ll be happy to be proven wrong and see it be more pro-Clinton (or anti-Trump) than the polls indicate.
The states that look solid for either candidate might experience low turnout for the predicted winner. It would be the states with an overwhelming majority favoring one of the candidates so it isn’t likely to change the electoral outcome.
The states that are traditionally assumed to be solid for one party are also the ones to get polled the least often, so yes, look there first. Then look at their recent trends in margins, particularly demographic changes, to see if the assumptions have been getting less valid. Then do your short-term predictions of the superficial trends over the next two months or so.
I’ll go with Texas. The Hispanic vote is an anti-Trump vote, and it’s massive.
A couple of other things about Texas that may make the numbers skew:
For obvious reasons, Ted Cruz is popular in this state and a lot of people like that he snubbed Trump.
For non-Cruz Repubs (and there are a lot of them), they are quite demoralized about national prospects.
Energized Dem base looking forward to GOTV efforts come early-voting.
For example, and I speak to the San Antonio experience myself, here is the local GOP branch’s website:
You cannot tell by anything in the front page that indicates that this year is a Presidential election year. In fact, you have to go to their Facebook page to find any mentions of Trump. (There are two Trump-oriented dates on their calendar, already passed.)
Hillary is prominently featured in the slideshow and on the front page. As is Bernie.
For those wanting to get in the weedshere’s a Pew study of the 2014 and the limitations of LV screens.
Of note of the LV models they tested in 2014 results ranged from predicting “a 2-point Democratic lead to a 7-point Republican advantage in the generic U.S. House vote.” They feel the benchmark should not be the actual result (which reflected some actual last minute decisions made) but the 3-point Republican lead among verified voters when they were interviewed prior to the elections. In that election past voting record improved upon other systems used. But different times may lead to different LV screen impacts.
Bottom line being that LV screening is a very inexact science and is not performed in the same manner by different pollsters which can lead to very different results from the exact same data set.
I think the former industrial north has the most potential for polls to be off. It is ground zero for resentful angry non-traditional-GOP white people. However, that doesn’t mean that all of them favor Trump, and more importantly, doesn’t mean that they will all go out and actually vote.
Without knowing the turnout of this demographic and how it can be changed by this Trump presidency, you can’t adjust the polls for “likely voters” because you don’t know how many disaffected whites will actually vote this time.
To a lesser extent this is also true for traditional GOP members who might be disgusted by Trump, who might just sit out the election, but I think that would affect all states fairly equally.
The reason for the thread is exactly that first part. We don’t know how turnout is going to be impacted by this particular contest so we have to make our guesses along with our guesses at how the LV screens will do at guessing them. And not just turnout of those two groups but every other one as well.
But different states have very different demographics. I’d be surprised if it impacted all fairly equally.
Trump’s campaign doesn’t seem to be putting much effort on local level campaign efforts, and has expressed skepticism of voter data-collection based campaigning. So it seems plausible the Dems will have a much better GOTV effort as compared to the GOP this year, throwing off likely voter models, which are based partly on the turn out in previous cycles.