[RD] US 2016 election: Poll watching thread

He could be wrong insomuch as his data smoothing is inaccurate, although yes, the election results won't by themselves demonstrate that. But we probably can and should try to decide if 538's model using weights (which are sadly not very transparent) or a more straight-shooter model like Princeton are more accurate for the future once we have more results.

Also you took my post way seriously than I thought anyone would lol, it was sort of silly but I wanted to do it because it seems like this site is entirely in the 538 model basket and there are plenty out there. Irrespective of which one finds better. I understand probability forecasts and "wrongness."

 

I mean people also regularly complain about how "inaccurate" meteorologists are. Hell I've seen people on this very board call for >90% odds on civ4 battles be auto-wins because their tank lost to a longbowman at 99%. People in general don't really understand how statistics and statistical models work. Educational systems should be better about that.

 

Sam Wang's model has a really low variance, low enough for me to just discard it as obviously overconfident. I'm with Cami on this one: anyone who thinks there's a 99% probability of Clinton winning this election doesn't understand on an intuitive level how probability works, no matter what their credentials and no matter if they even happen to call every state correctly in this particular election. I'm willing to believe Clinton's odds are higher than Silver's ~65%, based on early voting results in Nevada and North Carolina, but still well below 90%.

It reminds me of how Long-Term Capital Management had placed their bets in such a way that, according to their model, it would take an event 10 standard deviations away from the mean for their hedge fund to collapse; the associated probability is such that it would take longer than the age of the universe to have a 10-sigma event. Then the Asian financial crisis happened, and then Yeltsin's Russia defaulted, and then they collapsed ridiculously. LCTM was founded by a bunch of big-name people with PhDs in a variety of fields, but they weren't skeptical enough to realize that the >99.9999999% odds of success they were calculating didn't mean that this was actually an accurate depiction of reality, it just meant that their model was overconfident.

 

At this point, I'm taking 538 odds with a grain of salt. I'll read their articles, but the FBI release two days before basically puts Clinton ahead by two scores with 20 seconds left in the game.

 

I am going to concede Nevada on metalhead's logic. Early voting is too big a chunk of the total. Silver admits his model does not account for that. Given the new polls, it seems clear that the Trump push has stalled. Silver also said that a steady three point lead is a lot safer than a three points and falling. Three points seems high, but somewhere around two points is reasonable. The question becomes whether it is steady or yielding.

The Senate is another matter. Significant pushes in four states make me think that the GOP will hold on to 51 seats.

Of course, everything is close enough that turnout matters greatly. Several individual states have both close Presidential numbers and tight Senate racesThose could interact.

Not two scores, two points. It has not been two scores since the middle of October. The real question is who has the ball?

J

 

Which is significant for, among others, Nevada. The early Clinton vote is also probably an early Cortez Masto vote. Republicans should be less worried for Rubio's reelection because he's probably getting some of Clinton's hispanic vote.

NH is crazy on the other hand. The senate race polls have been all over the place for months, but right now the presidential polls are just as unreadable. Just today Emerson is giving Ayotte +3 and Clinton +1 while WMUR is giving Hassan +4 and Clinton +11.

 

Clinton was already ahead before the FBI announcement. The FBI release scored another for Clinton, which will never reflect in the pre-election polls. Pretty much all of Trump's TV smear ads depend on Clinton's emails, which she was just cleared for. The Trump campaign doesn't have time to re-do them.

She's up by two scores. Not even an act of terrorism will reverse that, because that would suppress regular-voter turnout and emphasize early voting. Which would be advantage Clinton. They both have the ball. Always did.

 

I'll play with you on this analogy cause it actually is relevant this time. Hillary has the ball and can just kneel down to run out the clock. The reason she has the ball is because the Comey thing is clearly an interception. Trump went all in on "ComeyisanAmericanherogodblessComeyHillaryisgoingtojail!!!wooohoo!!" That squeeze got picked off by the announcement which is being reported by both CNN and MSNBC as "HillaryhasbeentotesclearedofAAAAALLLLLwrongdoingOMGnothingcanstophernow!!!"

Of course Hillary supporters attitude is "told ya so, she is totally cleared of all wrongdoing whatsoever... nope, Email scandal finished, nope, not talking about it anymore, I said my piece and I'm counting to ten, it's over lalalalalalalalalala..." And Trumpers/TrumpSADs are left being drowned out, meekly gasping "but, but, but... Rigging! 'They' got to him! Everything is rigged! It's all a fix! Everything is corrupt! Everything is rigged against Trump! The refs cheated us!! 650,000 emails! Nobody could do that! It's a fix I tells ya!"

And that is how the election cycle appears to be ending... going into election day... Trumpers are tearing up their betting slips, throwing them to the ground and yelling "We was robbed!!!" as they watch Hillary take a knee and toss it to the ref with 25 seconds to go and no timeouts... While Team Hillary is saying "Nope, the ruling on the field stands... whew, that was close... now let's just take a knee and get the hell outta here".

So I ask you... Who is still watching/waiting around for the interviews and wrap-up (ie voting tomorrow)... and who is changing the channel in disgust (ie staying home tomorrow)? Who is saying "Well that was ugly... never shoulda been that close, but it sure was exciting! So who wants another round for the post-game wrap up?"... and who is saying "Ugh, that sucked... OK guys I gotta work tomorrow, *yawn* ...I'm headed out... see you next time."?

 

Interestingly, Harry Enten said that if you take Nevada off the board for Trump, the resulting polling error that would indicate correlated to other states would put Clinton's win probability at 88%.

My issue with Nate's model this time around is that it remains agnostic to the distribution of undecideds and third-party voters back to one of the two major-party candidates in the final vote counts, which puts an equal tail of uncertainty on both sides of the electorate. Earlier in the process, this is an understandable way to do it, because there isn't really any data to indicate where it might make sense to weight the uncertainty. But that isn't quite the case now, where we have both early voting and polling data that indicates which way these various buckets of voters are going. I feel like it's a poor choice to not to attempt to model the uncertainty, particularly since his probabilities can make big moves based on relatively small changes in polling margins. He simply cuts the undecideds right down the middle, which makes little sense in deep blue or deep red states especially where the undecideds will almost certainly break with the state's political leanings.

As he has noted innumerable times since 2012, polling is getting ever more difficult to carry out, ever less reliable, and actually is less numerous due to newspapers having reduced budgets to spend on such things. He has basically baked all of this into his model as uncertainty that breaks either way, but that feels like an unsatisfactory solution to me.

 

To be fair, it's not my logic on Nevada, it's Jon Ralston's. My logic is simply that when Jon Ralston says something about Nevada elections, you listen to him.

Closer to 3, since Clinton has gained a point in the 4-way, and 1.3 points in the 2-way RCP averages in the last few days. The top 5 polls on 538's national polling page currently all have Clinton +4. Ann Selzer has the national number at +3, and Monmouth at +6. For a variety of reasons I think +6 is closer to the final number than +3 or +4, but if you want to use the football analogy, Hillary has the lead and the ball, and is driving down the field to kick a meaningless last second field goal to cover the spread.

 

RCP has it roughly 47% - 44% in favor of Hillary, which leaves 9% undeclared. Applying the non-scientific handy-dandy Sommerswerd comprehensive fence-rider theory... I'm guessing with Hillary as the clear favourite, it breaks 7% to 2% Hillary's way, with Johnson and Stein getting negligible numbers... all in all, translating to roughly a 8% popular vote spread for Hillary.

 

Actually, I'm not even using a football analogy--more like a basketball analogy. An NBA team can come from behind by two scores in 20 seconds. It just doesn't happen 30% of the time.

 

I'm putting it at Clinton +7, 51-44, which is a tick or two lower than where I thought it was a couple of weeks ago, but still a plenty comfortable win. I think a double-digit win is plenty likely. A Trump win is exceedingly unlikely, though of course cannot be ruled out entirely. I think their negatives are such that minor candidates will get around 5% of the vote. She should get somewhere in the neighborhood of 350 EVs.

 

I see three sources of inaccuracy left in the polls:

1). Early voting. An inaccuracy which SHOULD be there. Early voting needs to be treated just like regular voting, where we don't know until the polls close Tuesday.

2). Media conflicts of interest. Who holds the polls? The media. Who stands to benefit to the tune of $100 million by political advertisers fighting it out in a close race? The media.

3). Unfair election practices and/or rigging, either from foreign sources or domestic. The pre-election polls are not subject to these influences; and with these practices being shady--by definition--it's hard to identify them in order to emulate them in the polls.

1) and 2) would result in a wider Clinton margin than the polls suggest. 3) goes for Trump, and for a very simple reason: you don't exactly see the white male vote getting suppressed.

 

So you are suggesting than the media try to present the race as closer than it is, so as to gain ads? Why? Hillary foundation will just pay the difference

Well done to those dems who helped cancel Bernie as the nomination. Cause more mud, partisanship, and muddy partisanship was what US desperately needed.

 

Just to be clear, that would not be a media pro-Trump bias, but a pro-underdog bias.

The Republicans can be as much to thank for Bernie not getting nominated as the Democrats. A Bernie-Trump contest would be a pretty cut-and-dried Commie Socialist against the billionaire businessman. That contest always goes to the billionaire businessman in the U.S..


edit: I accidentally left out a fourth inaccuracy factor: time lag. News developments in the last 5 days will never make it into the polls, but they still sway votes on Tuesday. That I also see going advantage Clinton. The news I see impacting election:

1) FBI clearing Clinton. The big one. p.s. the U.S. stock markets are already digesting this news, and they indicate a perceived Clinton victory.
2) Janet Reno dying. She's the first female attorney general, under Clinton.
3) War on ISIS in Mosul.
4) Rainy weather all throughout the Midwest. That decreases turnout. The big one is in Texas, which saw a strong Latino early vote.

 

^I don't think most people would see it that way. Although, yes, it would be presented this way by many media people and republicans - yet it is the stupid argument. I think Bern (despite being so old and not that charismatic) at least would be running a positive campaign, something the current nominees are not able to do due to being so crap.

 

In the fanfiction playing out in my head right now, tomorrow China launches a surprise cyber attack against the legions of Russian hackers that are preparing to interfere.

 

I don't think Bernie's campaign would be perceived as positive after Trump was through with it.