by Tony Crocker » Sat Aug 31, 2019 2:01 pm
I stumbled upon this old thread, which is certainly worth revisiting in view of 2016.
Nate Silver deserves credit for framing his election forecasts as probability distributions. He gave Trump a 29% chance of winning in 2016 and said there was a 10% chance Trump would lose the popular vote and win the electoral college. Usually there's only a 1-2% chance the popular and electoral college winners will not be the same. In 2016 the electoral college skew vs. the popular vote favored Trump over Clinton by 2.86%. Most national polls in 2016 were not that far off in the popular vote. Real Clear Politics, for example, predicted Clinton would win by 3.3%, but that the electoral college margin would be thin, with Trump winning Florida, Iowa and Ohio. Note that if Clinton wins the popular vote by 3.3% instead of 2.1% she carries Wisconsin, Pennsylvania and Michigan, resulting in the exact RCP call in both popular and electoral votes. Final national polls missing the actual popular vote result by 1.2% are not a big deal.
In 2012 the electoral college skew vs. the popular vote favored Obama over Romney by 1.61%. Therefore those polls which showed the popular vote a dead heat 9 days before the election still would have resulted in an electoral college win for Obama, with Romney adding only Florida and Ohio to the 24 states he won. Romney would have had to win the popular vote by that 1.61%, adding Virginia and Colorado, in order to win the electoral college. So it was probably reasonable for Nate Silver to say Obama had 65-70% odds at the point where the popular vote was even. And then when the late polls moved toward Obama those odds would gone even higher.
http://bestsnow.net
Ski Records
Season length: 21 months, Nov. 29, 2010 - July 2, 2012
Days in one year: 80 from Nov. 29, 2010 - Nov. 17, 2011
Season vertical: 1,610K in 2016-17
Season powder: 291K in 2011-12