I know that each and every one of us is all over the place when it comes to thinking about the outcome in November. For those who are on the optimistic side and would like a little bit of statistical support for their optimism, let me introduce you to this poll aggregating site, JHKFORCASTING.COM. Like 538.com and several other sites, it takes polls and sometimes adds some little bit of secret sauce I don’t understand and comes up with the odds of a Harris or Trump victory. It’s current odds, ranging around 58% or 59% tends to track nicely with 538, so I think this person‘s methodology in that regard is sound.
But the nice thing they add, if you poke around the page a little bit, is the ability to select a state, and then look at other states and their probability of voting the same way the index state votes. For example, if you assume that Pennsylvania will go blue, you can select Pennsylvania as your index state, and then click on all the other states in the country to see the probability that they too went blue.
As of this morning, if we were to assume that the blue wall was holding, and that Pennsylvania was blue, then there is a 69% chance that North Carolina will also go blue, and a 64% chance that Georgia will go blue and a 57% chance that Arizona will go blue. And if I remember my statistics correctly, that means there’s about a 44% chance that we get both North Carolina and Georgia and a 39% chance we win North Carolina and Arizona. But where it gets fun when we are feeling optimistic is if we change the anchor state. So instead of using Pennsylvania as our anchor state, let’s use North Carolina. Now here are several states and the probability that they will go blue if North Carolina goes blue. Georgia, 78%. Arizona 65%. And (insert drumroll here), Florida 52%
Now I can’t vouch for this site’s statistical math, but I can understand the logic being used, which is that the probability of different states behaving in different ways can be assessed by how other states have already behaved. So when people here start getting what sounds like a little bit psychotic with their optimistic estimations, there is a statistical basis to that optimism if one starts from an optimistic base.
Anyway, I suggest you go to the site and poke around, as it does have some interesting data that seems realistically connected to other poll aggregation sites and their predictions. In addition, if there are any statistical mavens out there, they can perhaps expand upon and provide a better explanation of how all of these contingent probabilities between states are statistically justified.
