One small update; Becca Siegel someone high up in the Biden campaign claimed this was the Biden’s campaigns final 2020 forecast in key states which is honestly eerily accurate. The x-axis is Biden’s share of the 2 party vote.
Notice the large variance in the variance of the distributions - apparently this is due to some states like GA and AZ having more helpful voter info stats than WI which it sounds like has almost nothing. Was a really interesting talk, and like how she said it’s not about what states can we win or do we need to win, but what states increase our odds of winning the most?
This is 100% correct. The reason for the differences in variance are because of better quality voter file data in some states than others, which makes individual voter modeling scores more precise in some states than others. The biggest factors differentiating states are party registration, race (especially if race is explicitly asked on voter registration forms, as in some southern states), and geographical polarization.
In states like CO, VA, AZ, and GA, you either have (quite predictive) party registration and/or race data for every individual voter, and also voters tend to be strongly geographically polarized, so it is easier to identify with greater certainty who is likely to be a D voter or a R voter.
Meanwhile in states like Wisconsin, pretty much all the voters are German/Scandinavian last name white people, without party registration, with geographical polarization being a lot less (fewer overly lopsided precincts and a lot more 40-60% precincts).
In addition, depending on
when the analysis is done, it is easier to predict with lower variance for states which have a lot of early voting, because then you already know explicitly which particular individual voters have voted for a larger share of the electorate, so you remove more of the uncertainty element from not knowing which particular individual voters will actually end up voting (turnout modeling scores are also reflected in the variance of those graphs, not just candidate support).
Another very important element here is that presidential campaigns (at least demographic ones) are well aware that when there are unexpected swings or late breaking events, they tend to swing some demographic groups more than others, and that voters with similar demographics tend to swing together one way or another across state lines. For this reason, presidential campaigns (at least democratic ones) try to give themselves multiple paths to victory which are more dependent and less dependent on different mixes of demographic groups.
For example, at the minimum you would like to have a "white people/midwest" path to victory available in case performance is strong with white voters, but also a "southern/southwestern/non-white" path to victory available in case. And on a more granular level, you would like to have a path to 270 that is more reliant on college educated white voters but also one that is more reliant on non-college educated white voters (more easily said than done for the latter at this point), and also distinct paths that are more dependent on black vs hispanic and/or hispanic vs black voters.
The reason for this is to try to avoid ending up in the situation of Al Gore 2000 or John Kerry 2004 where you are totally reliant on winning 1 particular state (Florida for Gore, Ohio for Kerry) in order to have a viable path to 270 electoral votes.
Now consider this again, and think about the reasons for contesting Florida:
like how she said it’s not about what states can we win or do we need to win, but what states increase our odds of winning the most?
This year in particular, one additional issue/demographic that might end up having a significant effect and "swinging together across state lines" is the demographic of women and the issue of abortion.
So one of the things that "increase our odds of winning the most" might be to contest states where abortion may (or may not) turn out to be a particularly important issue which could affect the percentage of the electorate that are women and democratic support, via the influence of ballot referendums. This provides an additional potential path to victory, a women/abortion path, which maybe (or maybe not) could hold up even if the other potential paths to 270 which are reliant on different demographics fail. Hence the reason to contest Florida.