And you are still doing nothing but guessing, there's nothing to stop the trend reversing over the next few days, but hey if you need to cling to the hopes of a Trump victory go right ahead.

To be fair, Don despises Trump,but was shell shocked by the Britex vote and seems not to grasp the differences between a referendum and a US Presidential Election.
 
The % chance of winning is a weird thing to discuss. Voters are not pachinko balls, the behaviour is not actually random.


Voters may not be random, but polls are. They're samples of voters, and sampling implies random sampling error. That's the fundamental source of uncertainty in the election forecast. Then you add in uncertainty from unknown events that occur between the forecast and the election that can change voters' minds; and then you add in uncertainty from not knowing who the undecideds will vote for. All these sources of uncertainty lead to a probabilistic forecast about the election results. I don't find it weird at all.
 
Just to give you guys an idea of how the 538 forecast moves, the following polls changed it by 1.2 points in Donald's favor.

Virgina - Hillary +12
North Carolina - Hillary +4
National - Hillary +6
Iowa - Tie
Georgia - Donald +1
Missouri - Donald +11

This doesn't make sense to me but whatever.
 
Just to give you guys an idea of how the 538 forecast moves, the following polls changed it by 1.2 points in Donald's favor.

Virgina - Hillary +12
North Carolina - Hillary +4
National - Hillary +6
Iowa - Tie
Georgia - Donald +1
Missouri - Donald +11

This doesn't make sense to me but whatever.

The differences in the map w.r.t. yesterday I note are that Ohio and Arizona have turned from light-blue to light-red.
 
Note that Silver's model is the only one below 90% (others are as high as 98% and 99%) and at this time in 2012 he only had Obama at 75%.

I'm not even a little bit worried.


I'm still worried. The only forecast I've seen at or above 98% is that of Sam Wang's so-called "Princeton Election Consortium." and I'm pretty sure his model is systematically over-confident: it is blind to the existence of undecided voters and wrongly ignores the correlations among state outcomes.
 
I'm still worried. The only forecast I've seen at or above 98% is that of Sam Wang's so-called "Princeton Election Consortium." and I'm pretty sure his model is systematically over-confident: it is blind to the existence of undecided voters and wrongly ignores the correlations among state outcomes.

The HuffPo forecast has Hillary at 98%. It takes into account undecided voters, though probably not to the extent that Silver's model does.

http://elections.huffingtonpost.com/2016/forecast/president
 
Last edited:
To be fair, Don despises Trump,but was shell shocked by the Britex vote and seems not to grasp the differences between a referendum and a US Presidential Election.

I don't think he wants Trump to win for ideological reasons but just so he can say, 'I told you so'.
 
Voters may not be random, but polls are. They're samples of voters, and sampling implies random sampling error. That's the fundamental source of uncertainty in the election forecast. Then you add in uncertainty from unknown events that occur between the forecast and the election that can change voters' minds; and then you add in uncertainty from not knowing who the undecideds will vote for. All these sources of uncertainty lead to a probabilistic forecast about the election results. I don't find it weird at all.

The weird part is that we think we are less uncertain about the uncertainty, that we have captured it somehow, reigned it in and measured it. It's the assumption that the uncertainty is of a certain type (and flavor) that gives us some certainty about it. :)
 
I really REALLY wish 538 had something on the House like they do on the senate. My attempt to Google this afternoon wound up leading me to an RCP link which said the Dems would take it. In 2006.
 
I really REALLY wish 538 had something on the House like they do on the senate. My attempt to Google this afternoon wound up leading me to an RCP link which said the Dems would take it. In 2006.
There is less polling data on the House. That said, Republicans will probably keep it, though I expect more people to vote for Democrats.
 
And you are still doing nothing but guessing, there's nothing to stop the trend reversing over the next few days, but hey if you need to cling to the hopes of a Trump victory go right ahead.
That's what everyone is doing. Sure, there's a certain level of educated guessing but still.
 
I wouldn't fret too much. Every aggregator has Clinton nearly locked into 272, with a high probability to get 330 to 350. The only map that is trying to give the GOP hope is the RCP average and they do that by giving credibility to Bob's Forecasting of Mizzoura, a pollster neither 538 nor PEC even bother to rate. That's how PA turned "toss-up" on their map.

Also bear in mind that if you toss out the hi/lo polls on a daily basis, Hillary's national lead is more like 7.5 than 5.7 currently shown on RCP.

If they pollsters who care about their ratings follow the usual pattern, they'll start moving to a realistic figure beginning of next week. They want to be within striking distance of the actual result on election night. (RCP doesn't care - they never take a poll of their own and can always deny their right-leaning choices to maintain that mask of neutrality.)

PA's the key. If it actually shifts to Trump, then Trump could have a chance in hell, but only if he sweeps every toss-up state, which is a huge order. All Hillary has to do is take FL or OH and the tally is back to a W. Or she takes NC and NV. But the chance of her losing PA is so small that even Nate's most conservative (of the aggregators) still has her at over 85%.
 
But the chance of her losing PA is so small that even Nate's most conservative (of the aggregators) still has her at over 85%.


Now down to 80% in Polls-Plus. One chance in five for Trump. Too high for my tastes. I hope his model is underconfident.
 
Now down to 80% in Polls-Plus. One chance in five for Trump. Too high for my tastes. I hope his model is underconfident.
He's down in every PA poll.

Silver's model is the outlier. Every other one has Hillary at at least a 94% favorite in PA.
 
He's down in every PA poll.

Silver's model is the outlier. Every other one has Hillary at at least a 94% favorite in PA.


He's the outlier, but he might be the only one doing it right. His model seems to take more seriously the possibility that the polls might have a consistent bias across states. As he explains here, in 2012 Obama outperformed his polls by 2 or 3 percentage points in almost every swing state. Were Trump to do that this election, he could beat the Electoral College forecast by 50 or more electoral votes, enough to win.
 
He's the outlier, but he might be the only one doing it right. His model seems to take more seriously the possibility that the polls might have a consistent bias across states. As he explains here, in 2012 Obama outperformed his polls by 2 or 3 percentage points in almost every swing state. Were Trump to do that this election, he could beat the Electoral College forecast by 50 or more electoral votes, enough to win.

Nah, even if Donald over performed his poll numbers by 2-3 points he would still lose. That might be enough to get Florida, Ohio, Iowa, North Carolina, Arizona, and Nevada, but he needs more than those to win.

And really, it is far more likely that he will underperform due to his terrible ground game.
 
Nah, even if Donald over performed his poll numbers by 2-3 points he would still lose. That might be enough to get Florida, Ohio, Iowa, North Carolina, Arizona, and Nevada, but he needs more than those to win.


538's polls-only forecast shows Clinton ahead, but by less than 3 percentage points, in FL (29 EV), NC (15), OH (18), AZ (11), NV (6), IA (6), and ME-2nd dist. (1). So, if the polls are systematically off by up to 3 points, it would seem that Trump could win the Electoral College.

And really, it is far more likely that he will underperform due to his terrible ground game.


I don't understand why the polls are systematically off in some election years, but you're mentioning only one possible one reason. Surely there are others.
 
Last edited:
538's polls-only forecast shows Clinton ahead, but by less than 3 percentage points, in FL (29 EV), NC (15), OH (18), AZ (11), NV (6), IA (6), and ME-2nd dist. (1). So, if the polls are systematically off by up to 3 points, it would seem that Trump could win the Electoral College.

Winning all of those states would not be enough to win him the election. He would have to flip at least one more blue state.

I don't understand why the polls are systematically off in some elections, but you're mentioning only one possible one reason. Surely there are others.

Kind of a big one. I will be surprised if Donald doesn't underperform.
 

Back
Top Bottom