2016 Postmortem
Related: About this forumThe trajectory of fivethirtyeights forecast looks now like...
Trump and Hillary's lines will Pass each other just before the election.
Which give the winning odds to Trump!
How the hell could Trump gain so many voters in the last few days?
I don't get it.
http://projects.fivethirtyeight.com/2016-election-forecast/?ex_cid=rrpromo
regnaD kciN
(26,044 posts)Although Trump has gained in recent days
Upshot lists odds of a Clinton victory at 85%
Huffington Post elections has it at 98.1%
DKos is at 92% (although that will probably change come morning)
Princeton Election Consortium gives odds of an HRC victory at between 98 - >99%
All of these (except DKos) update frequently as new data comes in; in PEC's case, it's updated hourly through the day, so this isn't a case of 538 working from more-current data, just that they have a different program to evaluate the data. In the opinion of Sam Wang (PEC), 538's program tends to "double-count" swings in the polls, and thus overstates them.
blue-wave
(4,352 posts)but I wouldn't get discouraged or panicked. There are certain factors here that these polls are not taking into consideration. Also, the polling can be biased toward the republicans. 538 even admits that there are issues with many polling companies, that's why they have a pollster rating page. Some polling agencies are simply better than others.
That said, I wouldn't underestimate the "Ick" factor that Trump brings into the game. There are many republicans that can't stand the guy, but would likely not tell someone over the phone their feelings. Also, in 2012, the race was allegedly tightening and many thought Romney had it in the bag. Remember Ohio and Carl Rove election night? It's for these reasons and more that I look at the recent polling with a skeptical eye.
We just gotta keep our eye on the prize and GOTV!!
sunonmars
(8,656 posts)pnwmom
(108,977 posts)this will be the time he finally has more support than she does? Look at the pattern. Sometimes she only had a slim lead, but it was always a lead.
But Professor Sam Wang at the Princeton Election Consortium has a theory. He thinks 538's projections are more volatile because he's double-counting state and national polls. Wang uses just state polls; but he recommends that people who are interested in an aggregate of national polls should look at the NY Times upshot, not 538.
election.princeton.edu
If you want an estimate that uses national polls, see The Upshot. Im a little concerned that FiveThirtyEights code double-counts (i.e. overcounts) the swings in national and state polls. Theyve been a bit underconfident and volatile.
still_one
(92,187 posts)do not represent significant numbers. People have already made their minds up, and that is not going to change.
This election is also unlike any other election. Previous models, such as 2012 do not apply. For one thing I do not believe that republican cross over votes are being factored in adequately. It is not a small thing when you have life long republicans such as kasich, bill crystal, david frum, who will not vote for trump. The republicans that make up that subset I don't believe are factored into these polls correctly. There was a Target Smart poll reported from Florida which represented votes from those who had already cast their votes, and it demonstrated a 28% cross over vote in Florida in those early votes. Let's assume that Target Smart is not a very reliable pollster, which it isn't. On a qualitative basis, the fact that it registered republican cross over votes means it is happening. The question remains is it significant? It is no secret that life-long republicans such as kasich, david frum, bill crystal and others will not be voting for trump, is another indicator that this election is unlike any other. My point being that I do not believe that the polls are adequately reflecting that cross over.
In these last days of the election, winning or losing has everything to do with getting out the vote. The ground game.
duffyduff
(3,251 posts)The crossover vote or Republicans simply leaving the top of the ballot blank or even a few voting third party will be unprecedented this year.
This is especially true with GOP women. They are not necessarily voting straight party ticket.
People completely underestimate how absolutely historic this election is.
still_one
(92,187 posts)DeminPennswoods
(15,285 posts)In the TargetSmart poll, Rubio is also significantly ahead of Murphy, 49-43 among early voters and 51-39 among likely voters. One would anticipate if early voters were not splitting tickets, the numbers for Murphy and Rubio would be closely aligned with the Clinton and Trump numbers. That's not the case. This data helps support the data that show 1/4 of Rs voting for Clinton.
still_one
(92,187 posts)are voting early, and in force. There is no question in my mind that a good number of these polls are underestimating women, people of color, and especially republican cross-over votes. That isn't Nate Silver's fault, he includes all polls in his evaluation, the less reliable with the more reliable. The problem is in this election the previous 2012 and other models that many of these pollsters are using do NOT apply. There IS a republican cross over vote, or republicans that refuse to vote for trump, and I suspect that is NOT being factored in properly.
It should be noted however, even with that, Nate Silver's model has the probability of Hillary winning is 66%, while trumps probability is 33%.
For one to assume that these probability lines before election day are going to cross in the other direction is not a valid assumption, because the polling is too volatile. It is based on models with many of these pollsters from 2012, and it does not apply. As I said, I do not think the republican cross-over vote is adequately being factored in. In critical states, the Latino vote is not factored in adequately, and perhaps the most telling is that the Democratic turnout vote is not factored in adequately.
However, even without those factors, those probability lines are NOT moving anymore for all essential purposes. The undecided at this time in the process are no longer significant. People know what they are going to do, and that isn't going to change.
blue-wave
(4,352 posts)There are at least three critical factors not considered in those polls. The huge surge in Latino registered voters in every state and as you point out, they are historically under counted. Latinos are very, very motivated this time.
The republican "Ick" factor. That is, the republicans, both women and men who can't stand Trump but won't tell a pollster over the phone how they feel. Early voting has republican cross over way up compared to past elections.
Lastly, some of these polling companies are just not very accurate. 538 has a pollster rating page for this very reason. We just need to GOTV!!
still_one
(92,187 posts)know what they are going to do. There are no undecided voters.
The effort now is getting out the vote, and that is where the Democrats have an advantage. The call banking and canvassing has been going on in full force, with emphasis on motivating people to vote early where they can.
That is going to make all the difference
Iskander
(12 posts)Remember Nate's origins he started out in sports betting. He also completely blew the primary with Trump. This is his him over correcting like a gambler that blew the bet he's become too risk adverse to throw it all at the sure thing.
Also his model doesn't take early voting into account and that same model is being fed garbage polls as the big names are currently busy doing in house work and not doing public polls so late in the cycle.
truthisfreedom
(23,146 posts)you might notice that it seems to be about to level off.
UCmeNdc
(9,600 posts)regnaD kciN
(26,044 posts)You're using 538 the exact way Nate Silver himself says you shouldn't use his (or any other research organization's) graphs: by visually "continuing the trajectory."
Each organization's numbers represent what they see as the FINAL odds, based on the current state of the race. While new data may change those odds as it comes in, it is a complete error to assume that "well, Hillary lost 1% of her margin since yesterday, so she'll lose another 1% today, and another tomorrow, and so on." Her margin could go up tomorrow (as some polls have shown). Or it might stay the same. What happened before isn't guaranteed to continue.
If there's any doubt of that, here's an exercise: go back to that 538 graph, pick any date a month or more ago, look at the direction of the graph for a few days before (where you have a consistent direction), and continue that "trajectory" into the future. Odds are, if you do so, you'll find that, by today, HRC should have had either a 0% or 100% chance of victory. Did that happen? More to the point, how many times did it even come close to the current figure? That's what happens when you try to "continue a trajectory" with these systems -- you fall into error by mistaking statistics-keeping with fortune-telling. If you insist on continuing a trajectory, the only proper way to do it is the simplest: as a straight horizontal line from the most recent data point to Election Day.
Grey Lemercier
(1,429 posts)Clinton 272 Trump 266 <<<<<< what Nate's EV map count yields at this exact minute
He is literally down to the outcome flipping to a Trump win if any, just one, of the following six states flip to the shitgibbon: CO, MI, WI, PA, NH, VA
its the nightmare scenario I posted (worked on it entirely before operation ratfuck) the same hour or so last Friday when Comey bollocksed the world
http://www.democraticunderground.com/12512552418
----------------------------------------------------------------------------------
atm, I think 538 is way off on some of the states
NV, NC, and FL to be exact, due to GOTV early results, especially with the Latino vote
jcgoldie
(11,631 posts)Even on Silver's lagging model, HRC is within tenths of a percentage point in NV, NC, FL and she even has a better chance of flipping AZ, OH, and IA from the red in your worst case map than Trump has of filpping any of the blue states you listed.
Demsrule86
(68,556 posts)Ralston's own lips...he is the expert on Nevada. 2/3 of the vote will be in.
Grey Lemercier
(1,429 posts)the Comey ratfuck of Clinton and will slide back down.
Let's also hope no last minute Rethuggery goes down.
Demsrule86
(68,556 posts)the damage they can do...and maybe Guliani will be sent to prison...hey a girl can dream!
jcgoldie
(11,631 posts)Clinton's odds on 538 have been hovering at around 67% for 3 days now. In fact they have actually rebounded a bit from a low of 64% on Wednesday. So your premise that the trendlines are on pace to converge is simply false.
FreeStateDemocrat
(2,654 posts)jcgoldie
(11,631 posts)Other than the nowcast model, the other 2 are supposed to project the finish based on current trends in the polls.
Grey Lemercier
(1,429 posts)Silver switched NV to Clinton earlier, is talking about NH possibly going red (gag)
Florida is going back and forth
wildeyed
(11,243 posts)NC dropped back into the blue. I looked carefully at that state's polling and noticed that there was on SurveyUSA outlier. I looked at the Senate forecast and the same poll was an outlier in that race too. So something weird with their sample, I guess.....
The election here is weird. My sense at my early voting site was that more Trump voters were voting than usual. It is a mixed area, with many immigrants, blacks and working class whites. When older, working class whites turn out at that location, I get worried. But there were also a number of immigrants signing up for same-day registration. And my strongly Republican MIL voted for Clinton, as did my daughter's friend's evangelical parents and a white Rep woman I know from college.
So who knows. It must be hard to poll. EVERYBODY need to GOTV!
getagrip_already
(14,742 posts)Don't worry over it. Even Nate is saying its off the rails at the moment.
duffyduff
(3,251 posts)It is totally different from any other one we have ever had.
budkin
(6,701 posts)She was at 64 yesterday, now 67. Even Nate tweeted yesterday things are looking better for her and the Comey "bomb" has run its course.
yellowcanine
(35,699 posts)I heard him last night and he thinks that Clinton may well get a couple of those states to go over 300. Personally I think she has a decent shot at all of them, depending on how much backlash there is about all of the FBI leaks to Giuliani and others.
LAS14
(13,783 posts)... is the opinion of the Clinton campaign for many weeks that the polls would tighten as the election approached AND that what they see pretty well matches what the projected.
yellowcanine
(35,699 posts)Deep breath folks. I think she has this.