RETVRN to 2020

The case for polls over forecasts, weird polling among seniors, the case of ActiVote, the Sanders-Trump nexus, post-debate polling shifts, Senate outlook

No. 329 | September 6th, 2024

🇺🇲 2024

Hopefully, I’m not stepping on this reporter’s toes, but in an interview this morning, I blurted out something clarifying: At this point, no election forecast is going to tell you more than a simple polling average in the battleground states. The fancy adjustments modelers make to account for “fundamentals” are doing more to swing the models than actual movement in the polls. And at this stage of the race, polls alone are what matters. To wit, this week’s averages — some of which show Harrismentum cooling off:  

• Silver Bulletin: Harris +3.0 (T+0.8)

• 538: Harris +3.1 (T+0.4)

• NYT Upshot: Harris +3.0 (H+0)

• The Hill/DDHQ: Harris +3.8 (H+0)

• RCP: Harris +1.8 (H+0)

• Cook Political: Harris +1.6 (H+0.4)

• VoteHub: Harris +3.5 (H+0.9)

• RacetotheWH: Harris +3.4 (H-0.2)

• Split Ticket (NEW): Harris +3.1 (-)

Average of the Averages: Harris +2.9 (H+0)

The race is very close when accounting for the Electoral College, though not quite a tie if you assume Electoral Bias will be the current 2.1 points. But as I wrote earlier this week, polls have tended to underestimate this number before, and if you assume it’s something closer to 2020, the race definitely is tied. 

The chaos of the last two months has just meant a RETVRN to the exact voting patterns we had in 2020, writes Nate Cohn: 

There’s no sign of the political chaos of the last few months. Instead, the results look typical: Nationwide, Kamala Harris leads Donald J. Trump by three percentage points, 49 percent to 46 percent. Across the battleground states, the race is a dead heat. In every state and nationwide, the polling average is within 1.5 points of the result of the 2020 presidential election.

In short, the polls finally show the close election that analysts expected a year ago, before President Biden’s candidacy went off the rails. If anything, it’s even closer than expected: The polling averages today are closer than the final pre-election polling in any presidential election in the era of modern polling — closer than 2000, 2004 or 2012, let alone 2016 or 2020.

What was weird about the 2024 election was that Biden’s candidacy at the age of 81 wasn’t exactly revving up younger nonwhite voters, and with him gone, some of the expected generational patterns have reasserted themselves: 

The Democratic lead among young voters is back. In high-quality polls over the last month, Vice President Harris leads Mr. Trump by an average of 20 points among the youngest reported demographic cohort (whether that be 18 to 29 or 18 to 34 in a given poll). The same polls showed Mr. Biden and Mr. Trump tied among young voters in July. Older voters, meanwhile, have barely edged at all toward Ms. Harris. Put it together, and the usual generational divide in American politics has returned.

(But the polling among seniors remains weird.)

Harris’ polling bounce was entirely before her convention.

Nate Silver on the challenge for forecasters. And he’s also out with something of a bear case for Harris in Pennsylvania. 

Our friends at JL Partners have launched their own forecast with The Daily Mail. It was far more sensitive to shifts in the race prior to dropping out, so it’s one to watch. 

538 has an explainer on ActiVote, the smartphone polling app with some of the spiciest crosstabs out there. Credit to them for reporting their unweighted sample, but somehow I don’t think a D+44 polling sample of Michigan weighted to D+25 is exactly representative. 

Split Ticket’s deep dive on Trump’s 2020 Latino support in Texas finds that a surge for Bernie Sanders in the primary predicted increases in Trump support.  

Polling after the first presidential debate (what counts this year?) shifts by an average of 2.8 points. 

Here’s the latest Senate forecast from Crystal Ball: