Nobody seems to be making much of it, but 538 is currently giving Trump almost exactly the same odds for a 2020 win as it did in 2016pic.twitter.com/cxwZ3kdLnQ
-
They are adding greater uncertainty based on distance from the election. Silver has stated that if the election were held today, it'd be like 90% Biden.
I'm starting to get the sense that nobody understands how models built on Monte Carlo simulations work. Obviously we'd be in a different situation if the election were tomorrow. The obvious fact is that it is not.
-
-
Yes, but it's a bit like apples and oranges in terms of interpretation. When you deweight structural variables the closer you get to an election, replacing them gradually with polling, you're effectively transitioning from one model to another.
- End of conversation
New conversation -
-
-
Uncertainty nets a wider spread of outcomes. But if you're running 40,000 simulations, the clusters will still be instructive. Can they tell you as much under high uncertainty? No. But if you run it again & assume tighter constraints for no good reason, you only confuse yourself.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
The diversity of counterfactual paths generated by a typical tractable model pales in comparison to the real world, especially when those paths are endogenous to agents in the system, and double especially when one of those agents is Trump, master of "how did we get here?"
-
It takes great effort to thoroughly model the diversity of even extremely stable, boring systems, and even then outliers are often ignored. Here we have a chaotic mix of heavy tailed dynamic processes being fed through a thresholding process (which is endogenous!) in *2020*
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.