Post-debate age concerns may have been the proximate cause of President Biden’s decision to step aside, but his lackluster polling mattered too. Donald Trump led by 4 points in the final update of our now-suspended Biden polling average, Nancy Pelosi cited the polls when trying to convince Biden to drop out, and poor internal polls may have been the proverbial straw that broke the president’s back.
It’s not just that Biden’s polls were bad — they were — but that he consistently lagged far behind Democratic Senate candidates. If you’ve been following polling this cycle (and if you read this newsletter you probably have been), results like this should look familiar:
These are weird results, at least compared to the last few elections. Split-ticket voting — voting for a presidential candidate and House or Senate candidate from different parties — has become really rare. It didn’t used to be. In 1972 for example, there were 17 split presidential/Senate outcomes. We saw similar numbers in 1976 (14 split outcomes), 1980 (13), and 1984 (17). Even in 2012, there were still 8 split outcomes. But fast forward to 2020 and there was just one split outcome. There were zero in 2016.
The recent Times/YouGov poll is a good example of how 2024 polls diverged from this trend: it showed Pennsylvania senator Bob Casey Jr. leading by 12 points but Biden trailing by 3 in the state.1 This means that at least 9 percent of respondents plan to vote for Casey but not Biden, and at least 6 percent plan to vote for Trump but not David McCormick (the Republican Senate candidate).2 This chart (stolen from Nate’s recent NYT article) does a good job of showing how consistent the pattern had been:3
So we know that Democratic Senate candidates were running ahead of Biden, but did his post-debate death spiral harm their performance? Nationally, Biden dropped by 2.2 points in our polling average between the debate and his exit from the race. His statewide performance was similar; our Pennsylvania polling average went from R +1.5 on the day of the debate to R +4.4 on the day Biden left the race.
We’d usually expect the president to drag Senate candidates down with him, but the performance disconnect held up post-debate. Take a look at this table of the average margin in 2024 Senate polls conducted before and after the debate. The Democratic candidate led in all six states with enough polls to be included before and after the debate. The margin in each state changed somewhat, but the change wasn’t uniform: Democrats gained ground in some states and lost it in others. For example, the Democratic Senate candidate gained 4.1 points after the debate but Democrats lost 3.1 points in Nevada.
Overall, Republican Senate candidates gained ground in two of the six races, and Democrats gained in the other four — although lots of these changes are pretty small, especially given the limited number of post-debate polls.
The lack of consistent movement contrasts with how Biden’s polls changed post-debate.4 Biden lost ground in 5 of the 6 states. His performance only improved in Michigan — by only 0.6 points — while Democratic Senate candidate Elissa Slotkin gained 4.1 points.
The difference in other states is even more striking. Trump gained 2.2 points in Pennsylvania post-debate, while McCormick (the Republican Senate candidate) lost 0.5 points. Nevada is the one state where the Republican Senate candidate gained more ground than Trump post-debate. But even there, the post-debate Senate average is D +5 while the presidential average is R +5.
So not only were Senate Democrats running well ahead of Biden, but his post-debate decline didn’t drag them down — at least not consistently. There are a few ways to interpret this disconnect. The most straightforward interpretation is that voters’ concerns about Biden were indeed age- and performance-related: Democrats had a Biden problem, not a Democrat problem, in other words. But there are other explanations, too. If voters expected Trump to win the election, they might have been voting for Democrats for Congress to preserve a balance of power. Or it might have been a case of expressive responding: people saying they wouldn’t vote for Biden to register their discontent with him, not because it’s what they actually planned to do. In this case, most Democrats would have “come home” before Election Day, and the actual presidential and Senate results would have been more similar than suggested by the polls.
If the disconnect wasn’t just expressive, we’d have seen an unprecedented level of split-ticket voting for a modern election come November. We obviously won’t be able to see how an election with Biden would have played out, robbing us of a sort of natural political science experiment. But I’ll be paying attention in the next month to see whether Kamala Harris’ polling continues this trend.
I just picked this poll as a recent example (and because I am somewhat of a YouGov fan) but the pattern is really consistent across public polls.
There are the results among likely voters.
It might seem low effort to screenshot an existing chart instead of making a new one, but trust me, getting past the NYT paywall took some work.
In both cases, this is just the margin from a simple average of the pre-debate and post-debate polls. No fancy polling average adjustments here. I’ve done this so the presidential and Senate averages are comparable, but it means that the presidential averages in this post don’t match our forecast polling averages exactly.
PA poll out today showing +2 Trump/+8 Casey. My thought is that a lot of voters are mad at the administration over inflation and other issues caused by the pandemic. If that's the case, Kamala will be in the same or similar spot as Biden. Dems should have run someone from outside the administration.
The disconnect between polls for President and Senate in swing states has an obvious interpretation which the NYT has not been pushing: Maybe this is a sign that political polarization in the United States has, after a brutal rise, finally gone into decline? Certainly when split-ticket voting became more rare, it was widely touted as evidence of increasing polarization. Anyone who agreed with that characterization (such as me) ought to consider whether 2024 polling is an early indicator that we have passed peak polarization. Woohoo! But maybe this lens is too optimistic to make for good journalism. :-p