FINAL PollFair Presidential Election Forecast @270toWin 🔴Trump 297 ✅ 🔵Harris 241 #GeneralElection #2024Elections #Trump #Harris #President #Map

Pre-election favorability ratings suffered from the same weighting issues that caused polls to be inaccurate in calling the election itself, which caused Harris/Walz to appear liked and Trump/Vance to appear disliked. You can see closer to reality in the exit polls.

It's looking like our election day model was pretty accurate, except for the Florida senate race which we called correctly but the margin was off. Florida PollFair Call: Trump 11 Actual Result: Trump 13 PollFair Call: Scott 8 Actual Result: Scott 13 Nevada PollFair Call: Trump 4 Actual Result: Trump 4 PollFair Call: Rosen 3 Actual Result: Rosen 2 Arizona (90% in) PollFair Call: Trump 6 Actual Result: Trump 7 PollFair Call: Gallego 1 Actual Result: Gallego 2 Our election day model is partially based on average polling crosstabs, providing more evidence that the poll surveys themselves were actually quite accurate and just the original poll weightings were wrong in many cases.

It looks like the average weighting from national polls ended up being just a fraction above D 0. Preliminary results from exit polls suggest the real environment was R 3. Toward the end, we saw the more accurate pollsters homing in on R 2, but during the whole campaign we saw anything ranging from R 2 to D 7. We find it hard to believe any pollster truly thought D 7 would be accurate. The interesting thing is the party-id crosstabs in most polls were close to accurate - if you apply an R 3 weighting you generally get results ranging from a tie to Trump 3 nationally from most polls, and our average would be Trump 2. So, there was nothing wrong with the surveys themselves in the main - it was almost entirely a weighting issue, as we suspected.

Having seen the success of our poll "Fairweighting" experiment, do you think that poll re-weighting is a valid exercise? Would you like to see more of it in future elections? Please leave your reasons/opinions/thoughts in the comments below.

State Forecasting Accuracy by Organization, Battleground States Only (Updated). Atlas Intel 100% PollFair 86% Rasmussen 86% >Quantus 86% Trafalgar 71% Insider Advantage 71% Real Clear Politics 71% >Patriot 71% Fabrizio 67% Remington 67% Emerson 57% >Redfield & Wilton 57% >RMG 57% Wall Street Journal 50% Beacon (Fox) 50% Marist 50% Morning Consult 50% >Activote 50% >SoCal/On Point 50% Suffolk 50% TIPP 43% 538 43% >YouGov/Times 43% Quinnipiac 40% CNN/SSRS 33% Bloomberg 33% GSG 33% Susquehanna 33% Echelon 33% Washington Post 33% >FocalData 29% Data for Progress 25% CBS 25% Siena 14% Cook 14% > = Added with update. We also wanted to add, we took data from almost all of these sources, even the ones that were 14% accurate. Our weighting adjustments fixed even the least accurate polls in many cases, and we managed to accurately call 6 out of 7 of the swing states.

The informal best and worst pollsters to work with this election season awards! (Note: this is not a list of the most accurate, but more about the most transparent and well organized, although they might also be some of the most accurate too once we finish our accuracy analysis). Best pollsters: 1) Emerson. What can we say - very consistent and presented all their data in a transparent and easy to consume manner. You can also tell they based their sampling and weightings around the 2020 exit polls. Weightings were very consistent and not constantly moving all over the place to compensate for crosstab movements (ie: no manipulation going on). 2) Atlas Intel. Their reports were a pleasure to work with. Everything presented in a clear and concise manner and also very consistent. Their movement in partisan weighting moved slowly over time and called the final electorate perfectly in many cases. We cannot say enough good things about these two pollsters. It was an absolute pleasure to work with their data and we weren't left scratching our heads about any of it even once. Worst pollsters: 1) Morning Consult. Data was a disaster to work with. Weightings constantly changing, things not adding up. More than a few very odd crosstabs (we won't elaborate too much on that). 2) Big Village. Again, things often didn't add up or looked odd. We thought about dropping them from our average more than once, but that would mean going back on our concept of accepting any poll with party id crosstabs. There were more bad pollsters than good in terms of working with their data, consistency and transparency. Everyone can learn a thing or two from Emerson for next time.

This was our most viewed post of the election cycle for whatever reason. We got a lot of flak for what we are doing, but ironically it looks like this Fairweighted poll might end up being dead on accurate - Baldwin 1. Read the comments if you need some entertainment this morning!

Thank you all for your incredible support this election! This has truly been a voyage of discovery we were on together, and we’re excited to share more analysis in the coming days. We are also pleased to announce we will be expanding our efforts for future elections. Your likes, follows, and sharing our page with others will help us grow and keep this effort going strong. We couldn’t have done it without you! Thanks again!

It's official! PollFair > Ann Selzer

👀

👀