Figuretti🇺🇦

TEXAS IPEV DAY 12 - WRAP Early voting is done. Final estimated early recall is Trump +9.7 among 2024 voters only. This reflects projected IPEV of Trump +11.2 plus an extra 1.5 of margin for mail. In 2024, this was Trump +14.9 for IPEV. @theoelections has it pretty close to that as well. The "shift" that kept shrinking was due to the 2024 baseline continually getting bluer over time - this cycle, that reversed - but during week 2 the recall stayed remarkably similar. Previous updates had estimated recalls that were undershooting Dems due to GOP enthusiasm in 2024, but this has been corrected by calibrating 2024 voters to the vote type on the canvasses which include granular election day/early splits. Overall, Democrats front-loaded their high propensity voters, and Republicans did the same thing a few days later. The last few days of EV got steadily less white though, and steadily lower-propensity. Among the 3/3 voters, the day 1 margin vs. the day 12 margin saw quite a drop - and this shows that pattern. The lean of the 3/3s, particularly among Hispanic voters, declined considerably. This makes sense, when you remember that in 2024, Latino voters disproportionately swung to Trump. Among lower propensity voters though, the decline was far more gradual, and Dems ended the period with a bluer pool of low-propensity voters than they did in 2024. In this case, their leans are red because many of the later-voting low propensity voters are minorities (day 11/12 had the fewest 3/3s of any day). There's an interesting effect in metros. As many of the other analysts have shown, blue precincts are turning out more than red ones, all else being equal. But *within* them, in the reddest precincts, Democrats are turning out slightly less than Republicans. This is why Tarrant County's final EV electorate ends up only 2 points bluer than it did in 2024. The non-2024 pool got steadily less white over time. In our 24 tracked counties, roughly 750k people voted who didn't in 2024. 30% of the 2024 voting pool was under 44 in our counties, while the non-2024 voting pool is 61% under 44. 2024 voters: 49% white, 28% Hispanic, 13% black 2024 non-voters: 34% white, 42% Hispanic, 13% black Who might these people vote for? The main thing that EV can tell you is not who will win, but who can't. This is not North Carolina or Georgia where the GOP is rolling over to die. The Republicans let Kansas wither on the vine to make Texas their Verdun, and have spent hundreds of millions to get their base voters out - so turnout alone is not going to be enough to forecast a D or R victory right now. The Democrats have taken the electorate from Trump +13.2 to Trump +9.7 roughly. Election Day will likely lean redder, but we don't know the size of the pool left to vote. Republicans are praying for gigantic turnout on Tuesday. The last few days have been encouraging for them - though the cumulative turnout is still really bad, they genuinely have had strong days of early voting particularly in the GOP base counties like Montgomery at the very end. That may carry to Tuesday. This election will be decided by 1) how much persuasion Democrats get in this electorate, and 2) how blue the non-2024 population is. The bluer the non-2024, the less persuasion is necessary. This is a Lean Democrat race. LEAN IS NOT SAFE. If the polls are too blue, Paxton wins. If actual persuasion isn't blue enough, Paxton wins. If there is an Election Day surge beyond anything seen in history, Paxton wins. But... if Texas polls are too red, Talarico wins by 4 or 5. Right now, as has been the case, Talarico is moderately favored, and I expect him to win by 1-2. The other statewide races all stay R barring a polling error undershooting Ds. Thank you all for following along with this project! It has been a wild ride - putting something together with a grand total of $500 in paying for county voter files and some ingenuity. Special thanks to @theoelections, @_fat_ugly_rat_, @DjsokeSpeaking, @ZacharyDonnini, @ADincgor, @lxeagle17, and @ChazNuttycombe for advice and talking through this.