Quote:
Originally Posted by JoltinJake
#1 is a clear yes. Proven by the graph in the link we were discussing earlier.
#2 I would say almost certainly not the same. I think you get diminishing returns once both candidates go below 50. We have n=1 on that right now, though, so no way to prove.
#3 I don't think would be predictive much at all.
So took me a while to come back to this since I was eating dinner.
The problem with models like this is that sometimes people look for patterns that prove their hypothesis ahead of seeing all the data or considering all the variables.
I agree with you on question 2 and 3.
If you believe 2 and 3 to be the case, then why do you say most definitely yes for question one?
If you look at the numbers, 2016 is the only year that neither candidate had higher than 50% favorability.
Would it be crazy to conjecture that in cases in which this is true, the favorability margin is not very predictive ?
That as both candidate's favorability approaches 0, win probability truly becomes 50-50 because you reach a Kang vs kodos situation.
Post electionwith additional data, we can see another anomaly. Why is trump able to get 40 some percent of the vote with just 30 some percent favorability ?
And therein lies the crux of the problem and can be supported by the many many more Trump supporters that state "well I don't like either, but trump at least promises me some benefit but Hilary doesn't speak to me at all" as compared to the very few like HastenDan who dislikes both but votes Hilary.
The problem is Hilary focused too much on lowering trump's favorability when it was already in the gutters instead of raising her own. Of course Comey didn't help. But it's largely because she didn't realize 50 vs 45 Is much more preferable to 40 vs 30 even though gap is bigger in 40 vs 30.
Erroneously reading stuff like the above 538 article only puts things in to a positive feed back loop.