Drifting Demographics: Diverging Directions in 2024
Is there any consensus in the latest national polls on where demographic groups are moving this election?
With election night right around the corner, it seems like there’s a new poll coming out every hour. Most show a close race. But if you dig into the crosstabs of these surveys and look at how demographic groups are sorting out, you find some interesting trends … especially when you compare current polling to the 2020 election.
I looked at nine recent national surveys with crosstab data: ABC News, Atlas, CES, CNN, Economist/YouGov, Emerson, Harris X, NY Times, and CBS News. I then compared these 2024 polls to the 2020 exit polls to see where Trump and Harris are estimated to improve on their 2020 performance among key demographic groups … and where this conflicting movement.
Strongest Consensus on Change from 2020
Latinos: Several polls show a significant decrease in Harris's advantage among Latinos compared to Biden in 2020. The average decline of -8.5 percentage points.
Blacks: Harris holds a strong lead among black voters (an average of 67 points), but her advantage has decreased by -5.0 percentage points compared to Biden in 2020.
Independents: These polls show a consistent and notable shift toward Trump among political independents. On average, Harris is performing -5.5 points below Biden’s 2020 performance. Harris has a narrow average lead of 2 points.
Voters 65 and older: There’s a steady shift toward Trump across these polls, with an average increase of +3.0 points for Trump compared to his 2020 performance among this age group.
Greatest Disagreement on Change from 2020
18-29-year-olds: Eight of the nine polls show Harris winning 18 to 29-year-olds (HarrisX shows a tie), but they vary on how much Harris will win this group. Some show Harris maintaining Biden's 2020 advantage, while others indicate a larger drop. The average change shows Harris performing -4.5 points below Biden’s 2020 performance. Still, there’s a large variance across these polls, from CBS News showing a +1 improvement … to Atlas showing -12 for Harris compared to Biden in 2020. See some of my earlier thoughts on this.
College Graduate: Harris maintains a 10-point lead over Trump among college graduates across these surveys. But there’s a wide variance. The average comparison shows Harris improving on Biden’s 2020 performance by +1.5 points. Some polls indicate little change like the Atlas survey), while others show a large increased advantage (7 points in the ABC poll).
Men: The consensus is that Trump will win male voters by 7 points. However, whether Trump will improve his 2020 performance is less clear. Averaging the polls against the 2020 exit poll shows Trump with a worse performance of -1.5 points compared to 2020.
Women: Seven out of eight polls show Harris winning female voters this election. The consensus seems to be that Harris will not perform at the same level as Biden did in 2020. The change for Harris among female voters averages -4.0 points, but the variation across polls indicates differing levels of movement … underperforming as much as 7 points or as little as 1 point.
The dominant media narrative seems to be that Harris will have a dominant showing among female voters while men are moving toward Trump. This data shows the opposite, at least in aggregate. Men are softening toward Trump, while women are less enthusiastic about Harris.
What Does It All Mean?
This post is partly a stake in the ground to see how things sort out on Tuesday. These results suggest that Harris will lose ground among some key constituencies, such as Latinos, Blacks, and political independents. Even small differences in how these groups vote compared to these polls could make a dramatic difference.
Even still, you'd think from looking at the top-line results of these polls, which ostensibly show a close race or Harris with a slight lead, that there would be more consensus “under the hood” on where these demographic groups are moving. But there isn’t. Harris leads female voters by 12 points in the NY Times/Sienna survey, while Trump holds a 1-point advantage among female voters in the Atlas survey. That’s a big difference. But NY Times/Sienna shows a tied race, and Atlas shows Trump leading nationally by a point.
My general suspicion is that pollsters are having difficulty getting people to answer surveys in an election where both candidates aren’t popular. Low-enthusiasm voters will determine the election. They’re also hard to survey. First, it’s hard to estimate if these voters will indeed vote … and hence be counted as a “likely voter” according to the survey. Second, they probably don’t want to answer questions about an election they aren’t excited about. That would mean low enthusiasm voters are either underrepresented in the survey or the sample is skewed toward people more willing to take surveys.
Even if the polls are correct and the election is close, there’s a possibility that many of these polls are wrong about many of these demographic groups. Having inaccuracies cancel each other out is not the same as being accurate. But we’ll know soon enough.