Here are our model’s estimates for this day in...
2020: Biden +9.6
2016: Clinton +6.1
2012: Obama +3.9
2008: Obama +2.4
2020: Biden +9.6
2016: Clinton +6.1
2012: Obama +3.9
2008: Obama +2.4
Why the difference? Our national numbers are informed by state-level trends, which didn’t show as huge a Clinton spike after the 2016 D convention. Our model also has a correction for partisan non-response, which shaves off a few points of her “bounce.”
Right now I’d say that it really, really matters if an aggregate adjusts for pollster effects (if it doesn’t, don’t look at it) and also if it uses state-level info (whoch provides more robust estimation). It’s a bonus if you get fancy like we do to adjust for phantom swings.
It seems silly to report a polling average if you know the swing in the polls is fake. That’s the idea behind our adjustment.
No, it just shows that things can change in 90 days
