There has already been much talk about how badly the polls “got it wrong” in 2020, just as there was in 2016...though in 2016, the polls did not get it all that wrong...the truth is a lot of states were VERY close, and most of the ones that were called wrong were still within the margin of error...they just came down, for our purposes, on the wrong side of the equation.
I tried, some of you may remember...to account for a lot of this bad polling, by introducing factors that people were more likely to respond truthfully to...and predictive about who they would actually vote for. At first glance, my final map looks pretty bad, but work with me here...
Looks like I blew it pretty bad, too...but I didn’t. I correctly called that PA and WI would be far closer than MI would be. I correctly called the close race in IA, assigning it red, as it did eventually go. I also got Arizona, and many will remember I kept pushing back on the Blue Texas folks, and got that right.
I blew Florida. And I blew Ohio. I incorrectly called Georgia...but it is very light red...and ended up very light blue instead...so I was not off much...and the calls of Ohio and Georgia in my map nearly cancel each other out, OH having 18 EV’s to Gerogia’s 16 EV.
I blew NC and ME-02. Note they are both very light colored...and as it ended up I just got the color wrong. I was within the margin of error on every call...even OH and FL...which I kinda blew. I think I did not weight quite enough for what I will call The Trump Effect...some may remember a thing called The Bradley Effect.
Anyway...the Bradley Effect has to do with race...whereas The Trump Effect...has to do with the fact of trump being a Cult Of Personality. I actually weighted that into my own projections. But apparently not quite enough. Without giving away my entire secret sauce, I weighted a few factors as how people answered certain questions...when I felt those questions would be predictive of how people would in fact cast their ballots. For example, people are more likely to answer honestly if they think the country is on the right track or the wrong track...than they are to answer honestly (or at all) “Trump or Biden?”
Obviously, if you think the country is on the wrong track, you are more likely to vote against the incumbent. I also baked into my secret sauce, factors such as the historical voting patterns of states — and i think I did not bake quite enough into it...I need to amend that sauce a little bit there...because some states went stronger Trump than I had predicted. It seems to me that those states :loyal to Republicans” are more strongly loyal to them...than states “loyal to Democrats” are to Democrats.
I’m wondering if...in both 2016 and 2020...if the national pollsters failed to fully account for this Trump Effect...even as I did...though not by as much as they did….because I did factor for it, just apparently not enough. And I wonder...for 2024, when a Trump will not be on the ticket...if there will be any residual effect or not. I guess we need to see what becomes of Trumpism in the GOP to answer that….but I’d think the effect at least will be lessened.
My own post-mortem would, initially, say that Biden just plain did not go negative enough...he’s too damned nice, if you want the truth. He should have been showing ads of Trump promising jobs, followed by workers leaving the Lordstown, OH GM plant. Or the Carrier plant in Indiana. All other issues aside, most people in the end DO vote with their own pocketbook...and we need to more-emphasize the fact that Republicans DO NOT bring prosperity to Main Street and Joe and Jane Sixpack.
There are those who will tell us we need to emphasize more...what we will do for WHITE people….as if this nation doesn’t already do enough for most white folk, right? But we do need to be careful to not give Republicans an opening to play on racial divisions by making us appear hostile to Whites….it’s a fine line. It’s a line I am not sure how we walk in the future...but Biden did a pretty good job walking that line, when he did speak of Scranton, and the blue-collar workers there.
Maybe we need to target our messaging better instead of targeting it nationally...we need to microtarget. It is said all politics are local, and I agree...even when you are talking about a national campaign. Because in the end, people are looking at “Who will make MY life better?”
We are a badly divided nation, as I see it...and Trump exacerbated those divisions. Will those in the GOP who follow him...do the same, or not? Will we unite more, divide more, or stay in limbo? I see Trump only as a symbol of what was already there, so I think The Trump Effect will be in play again, but to a lesser degree, in 2024...if only because there likely will not be a Trump on the ballot.
The polling industry, having twice got it wrong, need a serious autopsy. I think part of MY problem was my too-simplistic approach, as a result of the lack of time to dedicate to it...I don’t have crosstabs on my data...I relied on the data collectors to sample a truly representative sample of the nation, and I think they likely undersampled, once again, but less so than in 2016.
Were I to make a serious attempt, in 2024...I’d like to see a “control group” of, say 100,000 Americans from all walks of life, all geographic regions, all races, genders, education levels, etc...that reflects what our nation in fact IS...which can be readily obtained from Census data, which should still be reliably fresh. I’d like to factor that “control group” in...and see what swings exist in that control group...and measure that according to other polls...leavening somewhat, the swings, and possibly giving a more accurate picture when sampling a random group of people versus the same people.
In the end, every state I got wrong, was only barely so...and within the Margin Of Error...and instead of 353 EV’s Biden will get 306. Looking just at that number it looks like I blew it...but, digging deeper, I did not blow it by much...and I do not think the national polls did, either. But there are some additional factors that need to be more fully accounted for.