- Joined
- Jun 16, 2003
- Messages
- 24,414
- Reaction score
- 0
National. What about states?The polls were not off by 5%.
The final RCP discrepancy was Clinton + 3.2%. The actual final is going to be around Clinton + 1.2%. The polls were very good.
National. What about states?The polls were not off by 5%.
The final RCP discrepancy was Clinton + 3.2%. The actual final is going to be around Clinton + 1.2%. The polls were very good.
National. What about states?
National. What about states?
Actual Clinton lead up to 1.63%. RCP average was off by less than 1.6%. National polling was good.The polls were not off by 5%.
The final RCP discrepancy was Clinton + 3.2%. The actual final is going to be around Clinton + 1.2%. The polls were very good.
I have already offered a bet to anyone on this forum: If Trump runs again, he will win. No one has been willing to take it.I bet Donald Trump will be the greatest President of all time.
@Jack V Savage @m52nickerson @cooks1
Do you have any specific predictions about the Trump presidency? Anything that is measurable/quantifiable?
Define "twitter war".He will be the first President to get into a Twitter war.
Define "twitter war".
I find him to be rather predictable.No, but you know what I mean. No, I'm not going to bet on it. Trump is simply to unpredictable.
|RCP average - actual result|
PA: 3.1%
NC: 2.8%
FL: 1.1%
OH: 2.8%
MI: 3.7%
WI: 7.5%
NV: 3.2%
CO: 0%
NH: 0.4%
Taking the mean, you can say the polls were off by 2.73%. Only Wisconsin's polling error exceeded 5%.
Rustbelt polling was worse than other swing state polling, indicating that rustbelt LV models were worse.
Overall the polls were fine and your statement was inaccurate. The commentariat and "experts" like Sam Wang and Harry Enten gave Clinton a 99% chance of winning based on god knows what.
I will update this now since many states have since certified their results.
|RCP average - actual result|
PA: 3.0%
NC: 2.7%
FL: 0.9%
OH: 4.5%
MI: 3.6%
WI: 7.3%
NV: 3.2%
CO: 2.0%
NH: 0.3%
Mean state polling error: 3.06%
Mean state polling error excluding outlier Wisconsin: 2.525%
National polling error: 1.2% and shrinking.
Conclusions:
1. Bloomberg+IBD/TIPP+ ABC/WaPo national polls were very good.
2. Wisconsin state polling was horrible.
3. Other than Wisconsin, swing state polling was pretty good.
4. New Hampshire and Florida polling was excellent.
I have already offered a bet to anyone on this forum: If Trump runs again, he will win. No one has been willing to take it.
Hm. That's a good one. I'll have to think about it.How's this one? If Ivanka lives to be 65 she'll be elected President in her lifetime. Interested?
Wang's prediction was based off state level errors being less than about 2.5%. His model was off because the rust belt state polling was dramatically off (and in a biased manner).The commentariat and "experts" like Sam Wang and Harry Enten gave Clinton a 99% chance of winning based on god knows what.
Wang's prediction was based off state level errors being less than about 2.5%.
OH isn't really a swing state anymore. As for the rest, you asked how Wang's model got its 99%. It got that because it's meta-margin was 2.2%. I don't think Wang used RCP's averages so that's not the right comparison. Regardless, Wang was wrong so I'm not defending his methodology just mentioning why it seems to have been wrong (the polling data put in and the associate uncertainty).Incorrect.
Under that criterion, the only states that would be required to flip back to Clinton would be Wisconsin (RCP average Clinton +6.5%, actual Trump +0.8%) and Michigan (RCP average Clinton +3.4, actual Trump +0.2%).
Even then, Trump still gets 280 to Clinton's 258.
2.5% error in RCP average could have flipped PA to Trump. FL, NC and OH were already Trump territory according to RCP average.
If you replace 2.5% with 1%, you could make that argument, but you'd have to be a moron to build a model which assumes <1% error in state polling. I would wager we've never had an election with such a low swing state error.
The results of this election were very much in line with expectations generated by polling data. The "experts" are too arrogant.