Getting close to changing the polling average to my long used Presidential spreadsheet

Using cross tab demographic breakdowns from a large number of polls in order to basically create one super poll is how I got the 2016 election within 0.2% and how I managed to get the 2020 off by around a point, when pollsters were off by three-four points.

I am currently up to 15 polls where I have gotten usable crosstabs. This includes the recent Leger and Daily Kos polls which are large outliers showing Harris up by seven and four points respectively, so there might be a bit of bias still.

I believe I had closer to two dozen in 2020, when I decided to "go live" with it. I am waiting for a few more here before I feel comfortable changing over.

Interestingly, while there are many polls where my demographic modeling is off by a fairly significant amount from the top line, the over all averages is pretty close to identical. For example, remodeling the Leger poll by demographic breakdown shows me a 1.64% lead for Harris, rather than the 7% they show in their topline. Remodeling the Harvard Harris poll shows Trump with a 2.38% lead rather than the 4% lead they show. But overall the full modeling right now shows Trump ahead by 0.32% whereas the average top lines of those same polls shows him ahead by 0.25%. That is pretty close. So apparently some pollsters are still running odd demographics to show odd results, but when you average it all out... it might even out.

If these numbers look closer than the averages out there, I suspect it is because some of the crosstabs for polls favorable to Trump (such as Rasmussen) are behind paywalls. That being said, I am expecting at this point that popular vote will end up being closer for Harris than it would have been for Biden, simply because of her appeal to the ultra liberals on both coasts in those states where Democrats push the margins.  

Either way, once I get enough of these crosstabs, the top lines cease to matter much. If I was to look at where we are at today so far I see this:

• Trump is garnering nearly 88% of the Republican vote, 6% of the Democratic vote and 39% of the independent vote

• Harris is garnering 86% of the Democratic vote, 6% of the GOP vote and 34% of the independent vote.

88% percent of the GOP vote is exactly what Trump got in 2016, whereas he pushed that up to 94% in 2020. I suspect Trump will consolidate much of the undecided Republicans and come in closer to the 2020 number than the 2016 number. Clinton and Biden had similar results from their favored Party faithful (89% for Clinton and 94% for Biden). In 2016 Trump won the independent vote by about 4% while in 2020 Trump lost it to Biden by 13%. So he is actually doing better right now than he did in 2016, which of course he would have to if he expects to win the popular vote.