Premium

It is Trafalgar Versus the Polling World

AP Photo/Evan Vucci

In 2016, the polls were off…. by a lot. When I say a lot I mean several-times-outside-of-their-margin-of-error a lot.  In 2017, several of the polls admitted their 2016 faults and suggested their changes to make up for their 2016 mistakes, but yet, the methodology of most polls doesn’t reflect a change that would have fixed their 2016 mistakes.  Sure, things that would have fixed a problem here or there by 1 or 2 points, but nothing that would make up for 8 and 10 point swings by which some of these polls were off.

I decided to take several polling firms that had conducted polls in several of the battleground states and to put their predictions into a map as to what they think the result is going to be.  I was able to get maps together for New York Times/Siena, CBS/YouGov, Fox News, Rasmussen, and the underdog, Trafalgar.  Going from the less likely of the bunch, I will start with the Biden win polls.

New York Times / Siena

Let me start by saying, nooooope.  This is among the biggest misses you will see of the 2020 election.  To assume that Trump wins Texas, but loses other reliably red states like Iowa and Georgia is so far beyond anything that is logical or meets any of the other data available that it doesn’t make any sense. Not happening. I see Texas, Iowa, Ohio, and Georgia as kind of the canary in the mine for the Trump campaign.  If any of those states start going the race is over.

 

Fox News

Yes, you read that right, Fox News.  Fox News polling this election cycle has been an absolute disaster.  Again, you’d find that common voting behavior patterns would be found from one state to the next, but in the case of the Fox News polling, frankly, none of it makes any sense.   Why Texas would fall to the Dems, while Ohio would stay red is beyond me.  Not only did I find Fox News polling lazy this election cycle, but I also found like it was almost conducted with several different methodologies or turn-out models.  Fox News did not poll in either Florida or Iowa, but if Texas went, you can assume their same methodology would have had Iowa and Florida go too.  I never thought I’d say this, but Fox News polling is probably worse than New York Times’ wishcasting polls.

CBS / YouGov

I have voiced my opinion about CBS/ YouGov polls this season, I find these polls to be wildly unreliable in that they allow for opt-in, self-reported stats to weigh their polls, which leads to a lot of inaccuracies.  For instance, when it comes to the opt-in nature of their internet surveys, I think the likelihood of getting a legitimate Trump supporter to participate is remote at best.   What I will say about this map is that IF their sample was accurate (it isn’t) this would be about where the line would fall.  It appears to be an accurate “line” between the left and the right.  If you pushed the sampling of the poll to the left, you’d lose Ohio. If you pushed it to the right, you’d gain Georgia and Iowa. Still wrong, but at least logical. The methodology seems sound, but the sampling is off by a lot.

 

Rasmussen

Rasmussen was among the most accurate polling firms in 2016, however, this year’s map looks a bit bad for Trump.  To give a bit more context to this map, Nevada, Florida, and Ohio are all going to Biden by 1 point according to recent polling.  That is within the margin of error.  Rasmussen has not conducted polling in Minnesota, Iowa, or Georgia, so considering what states have fallen according to their methodology, I would put Iowa and Georgia in the Trump column and Minnesota in the Biden Column.  That would give Biden the win, 272 to 266.  I don’t believe that Trump doesn’t carry at least one of the rust belt, which according to the Rasmussen methodology would be Pennsylvania.  If Trump won PA too, he would win re-election.

Trafalgar

I think Trafalgar has been the most accurate of 2020.  This isn’t because of ideology or because “I want Trump to win.”  It is because if I were to conduct a poll of my own, Trafalgar’s methodology is as close to what I would do as they attempt to identify voting segments and means of best polling those demographic segments.  Their ability to identify and accurately poll the “shy” vote is what put them on the map, and I think it is about to do it again. People like to rip on Trafalgar, but you can’t argue with their record of success. Their map is as close to perfect at this time.  Trafalgar’s last Pennsylvania poll had a statistical tie, a loss of a 2 point lead for Biden from two weeks earlier.  If Trafalgar’s modeling and sampling are correct (and I believe it to be) the trend will continue to a Trump lead and eventual win of the state.

Conclusion

The problem with these polls is that the New York Times and Trafalgar can’t both be right.  Even if CBS/YouGov is correct, that makes the New York Times and Trafalgar both so laughably wrong that the entire polling industry will have to go into hiding.  If Trump loses, Rasmussen and Trafalgar go back to the drawing board for a few years.  If Trump wins, the collective that is media polling is over.  The Cook Politicals and Nate Silvers of the world can’t simultaneously claim expert status and get two elections so wildly wrong several times over.  Trump can’t simultaneously be polling in Ohio at 2% win and then win by 8.  The statements are just fundamentally incongruent.  While obviously, I’d like to see Trump reelected, I also have taken heat for saying certain states won’t go his way.   Liking a poll because it says what you want to hear, isn’t a good way for determining the accuracy.  Scoping out methodology, checking to see if the results fall within an ideological scale which you’d find the states (for instance a win in Michigan for Republicans but a loss in Texas), and making sure sampling meets both registration and other demographic identifiers, is key. Keep an eye on Rasmussen and Trafalgar polls in the next 24 hours.  Those are the ones that, in my opinion, will be most accurate come Tuesday night.