Based on the Early Voting numbers we're getting, it's actually dead on.
Electoral-Vote.com - FiveThirtyEight: An Assessment
- Thread starter brianmcdoogle
- Start date
- Status
Sohter.Nura
Member
This election is too much of a nail biter. Can I just wake up on Wednesday? My heart can't take this shit anymore.
God damnit Comey.
God damnit Comey.
ShironRedshift
Member
And considering the early vote results we have, that's nonsense. The results so far have been very favorable to D's in Florida and especially Nevada where they have a HUGE firewall at the moment. Nate's model is nonsense and much too reactive to things that aren't actually there.
hodayathink
Member
Doesn't early voting generally trend Democratic in the first place?
ShironRedshift
Member
Yes, but it's not as simple as that. Even accounting for that, the numbers we're seeing so far are crazy good. To explain, I'll quote this excellent post that does it much better than I could:
With those kinda numbers, it's just really hard to see how Trump wins states like Florida--it just doesn't seem feasible with the demographics so far breaking down like that.correct. I'm of the position that polls don't matter right now.
Polling is a lot more art than science, and much of that art is guessing "who is likely to show up on election day" then weighting your responses around those guesses. Many of the shit tier pollsters that lean R make some extremely questionable assumptions about minority turnout, and I've commented on that within the past few hours.
We have Early Vote numbers coming in, and just about all of them are saying that the assumptions about who was going to show up to polls and in what numbers aren't correct. Early voting numbers are completely obliterating 2012 participation levels, and not one single pollster saw that coming. none. Democrats are actually OUTVOTING republicans in Texas right now.
Think about this.
Have pollsters or aggregators actually corrected for this? Absolutely not. 538 was explicit about this in their chat today that they ignore early vote tallies in their model, and pollsters do too. So why are we still looking at polls which assume turnout numbers we know aren't accurate?
We can look at the states doing EV right now in significant numbers (no, PA and VA don't count- they're absentee but not EV) that haven't changed since 2012, and make some pretty accurate observations about where the race is.
Y2Kev name dropped Steve Schale- I actually like the guy. he's one of the reasons I'm so confident Florida is out of play- looking at EV data, not polls. here's a few observations:
From the votes we have coming in, we can see Florida's electorate is getting MORE diverse vs. 2012...not less. But we know this from immigration data. So Trump would need to carry a greater percentage of the white vote vs. mitt romney to carry Florida in 2016, because white voters are a smaller portion of the population. But let's continue.
The hispanic vote has surged substantially. through wednesday it was up 170K over 2012, and it's higher now. (someone posted today's figure somewhere, I can't be bothered to get it).
But the real story here is where they are. Many of the first time hispanic voters are those PR immigrants i like to go on about, though of course you have natural first time voters just aging into the system. Among hispanic republicans, 14% are first time voters, among democrats it's 26%, and among NPA it's 32%.
What this says is that the majority of those recent immigrants are registered democratic or NPA, and we know from polling they're all fairly hostile to the republican party by fairly lopsided 75-15 margins.
The NPA vote is also coming in MORE diverse than the state as a whole:
NPA is more diverse than the overall electorate, meaning it's going to be trending more democratic than the state vote overall, statistically.
also, from today:
Yeesh. both of those are TERRIBLE numbers for Trump. an electorate that is getting rapidly more diverse AND is female heavy overall is a disaster for him.
As for Nevada:
Those are very strong numbers, on top of strong numbers already in early voting. D's have amounted a huge firewall in the state. There's no way Trump is taking it unless something catastrophic happens.
Thanks for this post. The trends in early voting in Florida and Nevada have me feeling much better on how Tuesday is going to turn out. You get caught up in the horse race and feel stressed but in the end math doesn't lie.
Teren_Kanan
Member
I dont know guys. Living here in Florida I'm terrified. Fiance and I voted the other day. But there are Trump signs all over, and I constantly overhear people at my restaurant speaking pro Trump conversations.
Please dont let me down FL
Please dont let me down FL
ShironRedshift
Member
If you're nervous, you can always go down to your local party office and sign up for a shift of phonebanking or canvassing, if you don't already. If you've never done that before, they should provide you with a script to follow and everything you need. Of course, it's nonetheless very normal and common to be nervous about it even then, but once you get going that quickly fades. And it also helps to realize that if you do decide to say phonebank, they specifically focus on having you call people who have voted before and just remind them to show up to vote, so you shouldn't get many Trump supporters or anything. And either way, it's helpful to keep in mind that once the call is done you never have to talk to these people again if you don't want to. On the other hand, you're often able to meet some really cool people by helping out, especially at your local office--lots of cool people who are doing the best they can can almost always be found there.I dont know guys. Living here in Florida I'm terrified. Fiance and I voted the other day. But there are Trump signs all over, and I constantly overhear people at my restaurant speaking pro Trump conversations.
Please dont let me down FL
Every bit definitely helps, especially now, and being busy doing something helps to keep your mind off it, and it's just a great feeling in general.
Of course, I totally get if that type of thing just isn't your thing or you don't have time or whatever. Just a though.
But yeah, either way I totally get that and it's normal to feel that way. But the positive thing is those type of metrics are never reliable indicators of anything. Like, Romney and Romney supporters made lots of arguments based on stuff exactly like that and rally size and so on as undeniable signs that he had it in the bag, but we saw how that turned out. I'm not sure if that helps reassure you or not, but it's something.
Either way, wish you both well and take care! ^__^
EDIT:
There's also this:
Definitely a good sign for Hillary.
Morrigan Stark
Arrogant Smirk
I have a vague recollection of the full results being only available the next day last time, 'cause it takes so long to count all the votes or something. Am I wrong? Will we know the result on Tuesday evening, or on Wednesday? :/
KuwabaraTheMan
Banned
Both of Obama's wins were called by around 11 PM EST. The last elections to really drag on until early in the morning were the Bush elections.
I mean, Alaska and Hawaii won't be called until well after that, but the election will most likely be decided well before polls close out there.
Morrigan Stark
Arrogant Smirk
Cheers!
Aaron Strife
Banned
Hawaii's polls close at 11 est. I can't imagine any of the media outlets are going to be sitting with their thumbs up their ass waiting for results to start coming in before calling it for Clinton.
ShironRedshift
Member
More good news out of Florida:
EDIT:
And more from Nevada:
In other words, those are people that would not have been included in likely voter screens in the vast majority of polls. 82.5% of the Hispanic vote so far in the state aren't people who would have made it past the screens in polls. Very good sign for Clinton there.
EDIT:
And more from Nevada:
That's Vegas. Previous one-day record was like 48k or so, but it's already at 55k votes just from today with an hour left. And naturally Vegas leans very much D. Such strong turnout does not in any way match with Nevada being a tossup.
Basileus777
Member
I said this in Poligaf, but I can't really blame the models for not figuring out how to incorporate early votes into them yet. The real problem is that the punditry and writing that 538 has to accompany their model has really failed them. Silver is still pushing a narrative that belies much of the data available and it's fairly embarrassing. It's been obvious for almost a week now that Hillary has Nevada in the bag thanks to how the early vote is breaking, how is a site ostensibly about data analytics not incorporating this into their articles? Why are we getting low quality writing trying to justify a path for a Trump victory because of polling errors when there is early voting data to check them against?
perfectchaos007
Member
Wrong.
It is
Hillary Clinton
50.1%
Donald Trump
49.9%
Been that way for at least an hour.
Basileus777
Member
I mean 538 has been disappointing since its inception, but you'd at least hope their election writing would be up to par even if they've failed to fulfill the site's purpose of expanding data and statistical analysis into new areas. But they've been eclipsed by other sites even just in terms of political analysis of numbers.
I'm becoming more and more confident Nate's model is super wrong.
http://elections.huffingtonpost.com...residential-general-election-trump-vs-clinton
http://elections.huffingtonpost.com/pollster/2016-north-carolina-president-trump-vs-clinton
I don't know what he's doing, but it's obiously not lining up with aggregates and other models.
I don't care if he's "correct" on election day. I'm absolutely going to critisize him on his swingy, clickbait model come Tuesday.
This "follow the money" angle doesn't really do it for me. Surely Nate Silver could do 538 as a consultancy firm. Why bother doing bad math for money when you could go and do good math for money? Can't imagine he'd have trouble finding clients... unless he went and scuttled his reputation in a vain attempt to satisfy traffic pressure from ESPN? In which case he'd lose both the blog that depends on his reputation and do damage to a fallback career which also depends on his reputation.
HerbertBirdsfoot
Member
538 is baking in the possibility that all of the polling is skewed in some unknown way, and for one candidate. Sure, that's possible. And it's happened before.
At the end of the day though it's just running numbers in a model, and his model accounts for more uncertainty than some others.
We're seeing some really good indicators on early voting for Democratic counties and constituencies. We know damned well Clinton has a much better ground game. Lots of other decent analysis is available to suggest Clinton could outperform polling. But Nate's numbers machine turns a blind eye to all of it and still gives you a percentage change with a fricking decimal place in it!
I am worried about losing this thing, even if it was a 95% Clinton win probability. The thing about how would you feel about boarding a plane if you thought there was a 95% chance it would land safely. If Trump wins, we're in serious shit.
But come on there is no way Trump has a 30+% chance to win this thing when his path is so constrained. Come on, man!
At the end of the day though it's just running numbers in a model, and his model accounts for more uncertainty than some others.
We're seeing some really good indicators on early voting for Democratic counties and constituencies. We know damned well Clinton has a much better ground game. Lots of other decent analysis is available to suggest Clinton could outperform polling. But Nate's numbers machine turns a blind eye to all of it and still gives you a percentage change with a fricking decimal place in it!
I am worried about losing this thing, even if it was a 95% Clinton win probability. The thing about how would you feel about boarding a plane if you thought there was a 95% chance it would land safely. If Trump wins, we're in serious shit.
But come on there is no way Trump has a 30+% chance to win this thing when his path is so constrained. Come on, man!
Lord_Byron28
Member
Yeah I discussed this in another thread on here that it felt the media kept reporting a tight race for Trump and helping keep him competitive so he could survive the election. Just so they could have better ratings as Trump got people's attention.
KuwabaraTheMan
Banned
Wow, they really close the polls at 5 PM local time in Hawaii? I didn't remember that.
Yeah, that will be an instant call for Clinton of course.
Mike Works
Member
I agree that it's too much of a nail biter in that far too many people are voting for someone like Donald Trump.
But in terms of who's going to win, it's the least nail-bitey election in quite some time.
perfectchaos007
Member
IDK, the last one wasn't nail-bitey either. I think this election will end up closer than Obama/Romney. Obama beat Romney by 126 electoral votes. I think Hilary's MoV will be < 100 electoral votes
Ohio as well right? If trump can't win one of the 3 it's over.
If called for Clinton. Sure. But she is most likely losing Ohio.
Before election day, Obama was polling 2% less than Hillary was now. Most republicans were convinced Romney would win. It was a much more unsure election, even though Obama was still likely going to win.
There's a couple of things going on. The first is that there was always an expectation that the third parties would fade away and that the polls would tighten as we got closer to Election Day. What happens is that fence sitters or third party supporters will break away from them and towards one of the two major parties, and that this gain will generally make more of a difference towards the party that's trailing. News organizations will pick up this trend in terms of the horse race, so they say that it's getting harder to call. That's what happened in 2008 and 2012, so it's no surprise that it's happening this time as well. If things happened according to script, Clinton would win on Election day with relative ease, and outperform what most people think her poll numbers should indicate.
This time though, things are not proceeding according to script, and it looks like Hispanic voters are coming out in bigger numbers than ever before. What's more, a lot of them would have been filtered out by the Likely Voter screens so their numbers didn't end up factoring into the polls. If this is a significant occurrence rather than statistical noise, then anyone purely relying on the polls are going to be in for a big surprise.
Yeah, this election is going to be more lopsided than 2012. I never thought that Romney had much of a chance winning that one, and I've felt that Trump had even less of a chance this time round.
EatinOlives
Member
OK so I've been thinking more about the model and "uncertainty".
So as we know the closer we are to 50/50 chances the more the model is uncertain. And we know that there's apparently a lot of undecideds from Republicans that folded their vote back to Trump. But if that's the case, those voters are now in the Trump camp, therefore they're no longer undecideds, meaning the model shouldn't have that degree of uncertainty anymore. So what's bringing in the uncertainty now? It could only be explained if MORE previously solid people are back to becoming undecided, but that wouldn't make sense if the reported reason was that previously undecided voters are deciding to go back to Trump...
So as we know the closer we are to 50/50 chances the more the model is uncertain. And we know that there's apparently a lot of undecideds from Republicans that folded their vote back to Trump. But if that's the case, those voters are now in the Trump camp, therefore they're no longer undecideds, meaning the model shouldn't have that degree of uncertainty anymore. So what's bringing in the uncertainty now? It could only be explained if MORE previously solid people are back to becoming undecided, but that wouldn't make sense if the reported reason was that previously undecided voters are deciding to go back to Trump...
I don't think any of the polls actually check for and track people making decisions to vote. They just contain different people every so often (I think a few having rolling samples). It's especially difficult because they may decide one day for one candidiate, and the next day for someone else. That's why the polls often ask questions like "if the election was held today, would you be __<less/equal/more>__ likely to vote for <candidate>."
One of the biggest sources of the uncertainty, at least in the individual polls, is that they only poll a small number of people and try to generalize that (about 400- 3000 people). The margins of error offer a rough estimate of confidence about where the true means may be. There's a lot of other factors including bias and such that can introduce more variance and uncertainty.
EatinOlives
Member
That's fair, I was only wondering why the prediction is contracting closer and closer to 50/50, as if the uncertainty is increasing despite being mere days away from the election. I would think that the opposite would be true.
It's probably not totally fair to say the uncertainty is only increasing just as the two candidates are closer together. I think I was the one that helped seed that idea to you. It's a little more complicated than that, but generally the greater the spread between the two candidates, the more certain they can be (assuming they've done a reasonable job at the model and nothing totally unexpected occurs).
if we polled everyone, and they responded truthfully and never changed their votes, then yes as people made their decisions we would see less uncertainty. However, there's all sorts of tricky things going on. Often they poll "likely voters", which are deemed by some standards to have a high chance of actually going to the polls. This is important because not everyone goes out to vote, but they're willing to provide you with opinions over the phone. Depending on their demographics, they may also be weighted different too. So it may not simply be that 47% of the people polled in this poll were likely to vote for Bob. It might be adjusted to weight based on the demographics.
If you look at this graph, and check the "Undecided" box and move your mouse across the graph, you can see how the % of undecided voters fluctuates pretty heavily.
However, the biggest issue is that they're just asking small numbers and trying to generalize in a reasonable way, but these are just like educated guesses, based on a model generated from history and maybe some 'gut instincts'.
I'm not an expert on this, so I don't know all of the things their models account or don't account for, so some of this is just guessing.
maquiladora
Member
Better ground game, GOTV.
The USC / LA Times poll does:
http://cesrusc.org/election/
It has been (in)famous this cycle for showing large Trump leads.
I think you should divorce the notion of uncertainty as you think about it from statistical uncertainty.
Another sign that the media is really pushing the horse race theory was the constant story of low African American turnout in NC yesterday with early voting, in spite of the known fact that early voting was 1 week late in NC compared to 2012 thanks to the Republican legislature.
Personally, I thhink (hope) it's going to be a blowout due to how the "likely voter" model under represents minorities, and Hispanics in particular. I just wish that white educated males, alongside women who seem to be doing it, would show their brain's not broke and vote Clinton.
Or at least not vote Trump.
Because if you know what's up and still vote Trump,
Personally, I thhink (hope) it's going to be a blowout due to how the "likely voter" model under represents minorities, and Hispanics in particular. I just wish that white educated males, alongside women who seem to be doing it, would show their brain's not broke and vote Clinton.
Or at least not vote Trump.
Because if you know what's up and still vote Trump,
I just took a look at Ohio, and they have Hillary UP even though Trump won 4 of the last 5 polls (and 6 of the last 10). And out of those 10 polls, Hillary literally only won ONE POLL. How in the hell does that make sense?
This is dangerous confirmation bias....and I've seen it every where here. RCP and 538 were the go to sites when everyone talked about polling aggregates earlier this year. And now that they aren't reflecting the "Hillary will win by a landslide" narrative, people started trashing them.
BlastProcessing
Member
Well, this is an unusual election ... Nate Silver doesn't think his model is actually working well with it, himself:
http://www.rawstory.com/2016/11/fiv...-and-thats-according-to-its-guru-nate-silver/
That's a mis-characterization. He has said that this race is harder to predict than normal, which is why his model shows greater uncertainty. However, he believes that to be the correct way a model should react to a race like this, and believes that other models are too aggressive in their certainty.
A big problem with Nate Silver's model is that it's a poll based model that has a built in accounting for the possibility of a massive polling error... benefiting the Republicans. When the historical precedent suggests that there is indeed a massive polling error in certain states but it is actually in the other direction.
He has said many times that the model considers it equally likely that a polling error would favor Clinton. This is clear from the fact the model still shows a 5% chance that Clinton will win in a landslide (10%+ margin of victory) even though its average margin for her is 2.9%. Or you can just look at the distribution of projected outcomes on their page (it's a normal distribution).
The model also does adjust for historical accuracy of different polls. Historically Republican-leaning polls like Rasmussen or Gravis get adjusted in favor of Clinton. And polls with a poorer track record get less weighting.
Personally, I'd trust the other models mord and it isn't because of confirmation bias. They are better at statistics than Nate is.
If Hillary wins Florida does Trump have even the slightest chance?
- Status