Quote:
Originally Posted by Mathgod
The only error these polls made both in 2016 and 2020 was underestimate Trump's support. There are pockets of Trump supporters who never participate in polls, and are therefore never represented in the polling data.
There are differences between the rust belt states, for example a large Arab American population in Michigan, which is upset with Biden over the Gaza situation. Biden is also rapidly losing support in Pennsylvania among Black and Latino voters, as well as young voters.
So no I don't think all three states are joined at the hip. They are three separate contests.
|
Sorry I used the wrong word. I meant the Princeton Model. The Princeton model if you remember was forecasting a 98% probability of Hillary victory. This was because it assumed that a polling error in a state was an independent event and not a correlated event.
You made that error in your one in 8 comment. That isn’t the way swing state works. If you are down 3 points in Ohio and 2 points in Pensylvania and 1 point in Nevada and you win Ohio you likely win all 3. Essentially the odds end up being the odds of the largest polling error occurring and not the odds of 3 independent polling errors occurring.
I do agree that the under-sampling of Trump in the last two elections is a concern. I just think your 80/20 odds statement is not well founded with current data and the potential error in the current data.