View Single Post
Old 08-08-2016, 12:48 PM   #9935
octothorp
Franchise Player
 
octothorp's Avatar
 
Join Date: Oct 2002
Location: not lurking
Exp:
Default

Quote:
Originally Posted by peter12 View Post
Polls have increasingly fallen short of actual political outcomes. Based on recent experience in Scotland, GOP primaries, Brexit, the outlier turned out to be the reality.

I still don't think anyone in here really understands the mindset of the Trump voter.
I do think the difference is that the 538-style model has an absolutely phenomenal record of predicting electoral college results. It's possible these sorts of models get one or two states wrong, or miss the national result by a couple points. But in the past, the tendency has been for these models to miss on states that are low-polled because of an unlikelihood to affect the national outcome (either small population or only likely to flip in the event of a landslide national victory).
Even in the primaries, the polls were generally accurate at predicting the outcomes at well-polled state-level events, with the occasional big miss like the Democrats in Michigan. The problem on the Republican side was pundits (including Nate Silver by his own admission acting like a pundit) thinking that Donald Trump was somehow an exception whose front-runner polling status wouldn't hold. There's been nothing fundamentally different about Trump voters that makes the numbers less believable.
octothorp is offline  
The Following 2 Users Say Thank You to octothorp For This Useful Post: