October 30, 2020

My predictions have been wrong all along (slightly)

Two days ago I pointed out that if the polls in 2020 had 85% of the same October bias as did the polls in October 2016, Trump was leading Biden at 247 to 243 electoral college votes with 48 electoral college undecided. Today, with the same bias level,  and giving Virginia, New Mexico and all of Maine to Biden, Trump is leading Biden 264 to 260 with only New Hampshire and Minnesota remaining undecided.  

If Trump  gets Maine District 2 and wins New Hampshire, he actually gets to 269 - the dreaded tie scenario if Biden were to win Minnesota.  However I took a closer look at North Carolina which I was seeing as perilously close.  If Trump were to win that he would actually have enough electoral college votes to win the election even without New Hampshire and Main District 2.

It got me thinking - I'm applying both my own view of the 2016 bias as well as the 538 pollster bias. I could be double counting the bias.  But thanks to a video (I cannot relocate the exact one) from Red Eagle Politics (a great YouTube channel worth checking out), I decided to take a closer look at the 538 pollster bias ratings. As was pointed out on the video I saw, the pollster ratings provided by 538 are kind of suspect. The New York Times/Siena polls are rated as a +0.7% bias towards - Republicans??? 

Looking Just at North Carolina polls in October 2016, they had Clinton +2 (late September) and Clinton +8 (late October).  In their poll of the last three days before the election they rated it a 44-44 tie, with a margin of error +/-3.5% (their CYA poll).  In North Carolina in 2016 Trump won 49.8 to 46.2.  That represents a victory of 3.6%, actually just outside their margin of error. And in fact, they underrepresented the Trump support by 5.8% and Clinton by 2.2%.  Clearly (1) they got it wrong by even their own Margin of Error on 1 of the 2 data points as well as the overall result, and (2) how the hell are they rated Republican +0.7%? What is that rating based on? Perhaps congressional polling results? It doesn't matter, their bias factor clearly cannot be  attributable to upcoming election at a presidential level. That's just one example, and I found others.

The takeaway - The 538 pollster bias is sketchy at best, and I am backing it out of my calculations.

Stay tuned for a revised update of my polling-adjusted view of the electoral college situation later today. That's not to say all of the 538 data is bunk.  The do provide a pollster grade rating that is, for the most part, reasonable - especially when it comes to lesser known pollsters.  I will still be filtering out those pollsters that they have rated below a C+.  That may be a generous inclusion of me when it comes to certain pollsters, but I do still have other means of filtering out garbage polls (old polls, Margin of error issues, sample sizes to small for example) and those will remain intact. 

I expect my look later today will be more accurate than what I have been showing and I expect it to break slightly more in Trump's favor.

No comments:

Post a Comment

Disagreement is always welcome. Please remain civil. Vulgar or disrespectful comments towards anyone will be removed.

Related Posts Plugin for WordPress, Blogger...

Share This