Reply With Quote
She better get the blacks out
Silver is projecting 289 EV currently for Clinton in the polls+forecast model.
But that's not a 289EV map. It's a 273EV one.
If there is another BREXIT level minor polling error in the GOP's favor then New Hampshire is deciding the presidency if this map holds.
She better get the blacks out
Silver also pointed out earlier that polls do a terrible job of sampling the Hispanic vote, and Hispanics are coming out in droves in EV.
We'll see.
the lowest% blue state there is New Hampshire at 60.6%
among low% red states you have:
Nevada (50.4%)
North Carolina (53.7%)
Florida (53.3%)
Market Indicator Gives Trump An 86% Chance Of Winning The Election
http://www.zerohedge.com/news/2016-1...nning-election
Author: Tyler Durden
I'm a big fan, but Nate Silver's model is broken this time. Or if its not broken, its not equipped to handle this election. If Trump is really at the odds he's at on 538, then the prediction markets sure aren't buying it (neither am I). They are all substantially higher in price for a Clinton win than 538 is. The Upshot Group at the NY Times has odds that are much more favorable to Clinton but simply picking them because they have better odds would be confirmation bias.
Today there was a poll that showed Clinton as +18 in Virginia. The effect that poll had on the model? +1% to Trumps chances. How in the does a poll that shows +18 to Clilnton in Virginia (which frankly is flat out wrong) move the needle in Trumps direction? There have been movements like this a lot lately. I know he has area corroleation and the trend lines matter A LOT in his model, but this just makes no sense. There have been a lot of instances where polls that do NOTHING to change the electoral map do a lot to move his model. Its possible he's just overthought things and made it too complicated this time around, but its also possible everyone else is wrong and he's right.
In any event, I think Silver's model is dead wrong on Florida, NC, and Nevada. I really feel as those states are going to go Blue easily because of their early voting numbers. In Florida and Nevada this is due to Hispanics. We'll see what the final margins are and how they compare with Silvers predictions, but I suspect this will be by far his worst presidential predictions.
Have we ever dissected Silver's numbers like this though? I can't remember if he had this kind of tracking in 2012. We're looking at percent chances of a victory, but ultimately he's going to predict whoever has the highest % for each state and the total electoral result.
I do think the unprecedented Hispanic vote this year is going to be the surprise vs. polls though.
He had state by state numbers in 2012 and 2008. The difference is that his model has gotten way more complex. So now national polls are effecting the race when state polls maybe saying something opposite. To put it simply, I think his model is too complex for its own good and now in a race where the polls have a pretty damn wide spread and more voting is happening early than ever before its going to hurt him. Statistically I'm sure most of the states will end up within his mode's MoE but I just don't buy that Clinton is only a 2:1 favorite in a race where she's probably 3:1 to win the most important state: Florida.
VA, NH I think will go Trump. But i'm not sure he has FL locked down. If he doesn't get FL it's doesn't matter what else he flips.
I do think the unprecedented Hispanic vote this year is going to be the surprise vs. polls though.
...One early poll says that 28% of GOP voters are voting Clinton, but the Hispanic vote will go heavy Clinton...
Silver's model assumes significant correlation between states. So if Hillary drops a point in MI, even though her MI margin might still be comfortable, he assumes that she is likely to drop in states like PA, OH, WI. The model also factors in trend lines. If her numbers show a gradual decline over a couple of weeks, the model amplifies this because of the negative trend.
Conversely, if things move positively for her, the model boosts her chances significantly. So it's got an inbuilt virtuous / vicious cycle effect.
In reality, I think it's less likely that people flip their choice from week to week based on new information. It is more likely that people's willingness to respond to polls goes up and down based on the news cycle. After Pussygate, a few Trump-leaners might have declined to answer a telephone or online poll. Similarly, after the FBI letter, less Dems would have been eager to respond to polls. Respondents marking themselves as "undecided" are likely turned off by the show, rather than waiting for something new to evaluate - I mean, how can you still be undecided about these 2 candidates after the amount of that's turned up?
I don't think the fundamentals of the race have changed as much as the polls indicate. I still think it's a Clinton + 3, 300 EVs type election.
its not that undecideds aren't familiar with the candidates by now. my assumption is that many of the undecideds dont really care much for policy/nuance but are literally voting for who they think its a better person, or some other stupid reasoning along the lines of the "who would you rather have a beer with"
anybody who actually has reasonably strong political views knew who they were voting for months ago
Silver went off on a twitter rant today directed at the author of an article on HuffPo that (probably a bit too unfairly) criticized him for his model's odds. Nate's never claimed his model did anything but predict based on the polls, so obviously he doesn't factor in early voting numbers or anything else of the sort. But i felt his tweets weren't totally honest either. For one, the idea that his model has a precision that makes reporting chances to the tenth of a percent is ing laughable. Yet he does so, for entertainment purposes or whatever else drives his website traffic. Furthermore, he loves to talk about how his model is emperical and built off of evidence, but his model isn't public and it now does a lot more adjusting for variables as Warlord pointed out. While these items might be correlated, they're not perfectly understood and changes in the actual reasons driving the correlation might have changed since 2012 which alters the true statistical relationship and thus makes their use less appropriate. So saying that they're proven is fine but that doesn't mean that they're being used correctly for this election.
Still a fan, but Silver is passing himself off as way too above the board right now. There are plenty of people out there who have more statistical knowledge than Nate - Dr. Sam Wang for instance - and disagree.
Actually I think a major swaying point is who do you want to see less on TV during your normally scheduled programming? Do you want to see Hillary giving the State of the Union address? Do you want to see her walking across the Whitehouse lawn getting onto Marine One and throwing out a eyed salute? Neither candidate gives good face. Donald is a sourpuss who I cannot even imagine being cordial or diplomatic, and Hillary wears "we are better than you" all over her face, a bit like GWB did.
IIRC, Nate admits his model is quite conservative, and gives Trump better odds than most other models, tbh...
The comments about Hispanics makes sense too, they're 25% of the population in Nevada, and a huge chunk in Florida... their turnout has always been generally low, but if a decent amount does show up to vote (and I think this time around they do have an incentive to do that), then that will likely be the nail in the coffin for Donald, tbh...
wishcasting?
I think there will be a measurable percentage of undecideds that didn't want to admit they were voting for Trump.
I imagine the same will hold for Clinton.
cherry-picked?
where do you get your info, Dirk?
I think that map is correct on every state with the exception of Nevada and New Hampshire. All of the early voting data says Clinton has a huge edge there and Silver admits that that isn't being factored in, while the polling shows Trump has all the momentum is that NH.
NC and FL are close but I think the black vote has a lackluster enough turnout to give both states to Trump (early voting shows this is definitely the case in NC). Only thing that might change Florida is if the Cuban vote randomly swings left which isn't likely to happen.
IMO the race comes down to whether or not Clinton holds PA. That's a state that could have a lot of silent Trump supporters because of his views on trade while it also has an energized base in central PA.
There are currently 1 users browsing this thread. (0 members and 1 guests)
Posting Permissions
