WPAi Weekly Update – August 30

A quick note this week, with everyone (hopefully) taking their last mental break before Election Day. 

While polling often overrepresents a third-party candidates’ support, RFK Jr’s withdrawal from the race and endorsement of Donald Trump will have some reverberations. First and foremost, it is worth bearing in mind how few votes might be needed to swing the result in some of the battleground states. Arizona and Georgia were decided by less than 12,000 votes in 2020, and Wisconsin has been decided by less than 23,000 in each of the last two presidential elections. It might not take much to have a big impact. 

RFK Jr already saw a drop off in the polls after Biden exited the race. He was the landing spot for many of the “double haters” for Trump and Biden, with the Washington Post-Schar poll finding that 53% of double haters would vote for Kennedy back in June.  

Kennedy’s supporters trended both younger and less likely to vote. They also reported themselves much less strongly attached to Kennedy than other candidate’s voters, with 18% saying they back Kennedy strongly compared to 64% for Trump and 62% for Harris. The numbers for “extremely motivated to vote” were 23% for Kennedy supporters, as opposed to 72% for Trump supporters and 70% for Harris supporters. In short, the small Kennedy percentages in battleground states probably overstate his actual impact on the popular vote – but the margins are so small that it may matter. 

As an added complication, it looks like RFK will not be able to remove himself from the ballot in Wisconsin and Michigan, and at least some North Carolina counties have already sent ballots to print with RFK on them. Will he draw many votes given his withdrawal and endorsement of Trump? Probably not, but it is an additional marginal effect the Trump campaign would rather not have to deal with. 

While RFK’s withdrawal will have some marginal impacts, one of the fundamentals of this election is now Kamala Harris’ favorability. Whether or not Harris sees much of a “bump” in her ballot position post-convention, she has certainly seen an improved image among the electorate since July. In their final poll before Harris became the Democrat’s (presumptive) nominee, Quinnipiac had her image at a net –14 favorable to unfavorable (37% to 51%), with 11% saying they had not heard enough about her to have an opinion. This week, they released a survey showing her with an even image: 47% favorable, 47% unfavorable with only 5% not having heard much about her.  

Harris has been a deeply unpopular Vice President, so where did that shift come from? Harris had the biggest improvement among voters 18-34 who went from only 20% favorable to 52% favorable. Her image also significantly improved among Independents (+16 favorable), Democrats (+15), and women (+13). Harris’ favorable numbers improved across racial lines roughly equally (+10 with white voters, +8 with Black voters, and +11 with Hispanics). But clearly, younger voters are connecting with a candidate much nearer to their age than Biden. 

One interesting effect is that rural crosstabs (and thanks to Split Ticket for compiling this material) on public polls since August 23 shows a D +12 shift relative to 2020 vote. Is that real or a sampling artifact? We can’t be sure yet, but it might be impactful in Wisconsin and Pennsylvania, among others, if it holds. We’ll be digging into this one next week! 

Finally, we’d like to point everyone to a nice little resource on public polling by our compatriots over at Pew (Thanks for Scott and Courtney for doing this!). Note in particular the proliferation in contact methods to generate interviews, and the error types not accounted for in the ‘margin of error’ –  which only refers to sampling error. All research is an imperfect representation of the real world, so it is always helpful to be alert to the strengths and weaknesses of the tools you’re using. Cross-reference and triangulate whenever possible to gain the most wholistic view you can! This is a good resource to share with your friends and colleagues who may not be as familiar with the polling world.