I admit it, I peek into the polls occasionally. Well, maybe more than occasionally. Depends on what you think “occasional” means. Maybe every three days?
Up til now, I haven’t confessed my guilty secret. But today, I learned a couple of things which I must must must reveal.
538 is still my place, mainly because they give you all the individual pollsters’ numbers, even the ones I haven’t heard of previously, and tell us who commissioned the poll.
So today I checked in but instead of going to the actual polling for the presidential, Senate and House races (I check a lot of them), I went into something 538 offered — a make-your-own US map with each state’s electoral college numbers. That is, you start with a map on which the states are all pale tan or something and then you click on the states, making them blue or red, depending on how you think the election will go.
Above the map is a linear graph with Harris on the left and Trump on the right.
I was entirely pragmatic here: I made the solid red states red and the solid blue states blue. No fooling around. As I clicked, the electoral numbers were transferred up to the line and added in. (It still beats the band to click on blue states’ huge numbers and red states tiny ones. No matter how many red states you create here, the total number doesn’t rise by much.)
I gave Harris Arizona, Georgia and Pennsylvania and all the big north states, but not North Carolina or Florida. That is, I was rationally optimistic, following Biden’s 2020 results in Georgia and Arizona. I even divided up Maine’s numbers and Nebraska’s, as both states do.
At optimum, Harris had more than 300 electoral college votes.
Then I made Arizona and Georgia red.
Harris had safely more than 270.
No matter how I worked that map — worse for Harris and better for Trump — Harris was over 270. Does this mean we shouldn’t get too agitated on election night? I think so.
◊
Now, the deeply interesting info.
538 has a chart regarding the state-by-state accuracy/error rate of polling. I glanced at the error rate numbers dating from 2016. And here’s where I was startled.
Below, I’ve pulled the numbers of polls per state and then the error rates. I stuck mostly to states which usually have tight races, but also looked at red states where there are important tight Senate races, like Ohio and Montana.
You will see two numbers after each state name; the first number is the number of polls that are done in that state, and the second is the error rate of the polling. I’ve highlighted the most striking error rates.
In combination with the credible analysis that polling this year might be especially unreliable because Trump values polling more than votes, and his campaign has “bought” some polling results, maybe to keep him happy. So buckle up:
N. Carolina 157 – 4.4
Arizona 162 – 3.7
Ohio 77 – 5.4
Georgia 188 – 3.4
Missouri 60 – 5.5
Florida 204 – 5.0
Penn 147 – 4.5
Wisconsin 81 – 5.6
Texas 66 – 4.0
Iowa 51 – 6.7
Montana 27 – 7.7
Tenn 23 – 7.6
Idaho 9 – 11.1
Okla 27 – 10.4
S. Dakota 20 – 11.3