We’ll go way out on a limb here, and say that 2025 does not appear to be a good year for making political predictions. But, if anybody can do it credibly, it would be Kyle Kondik, who writes at Sabato’s Crystal Ball:
1. The electorate will be smaller.
Midterm electorates are not as big as presidential electorates, and there is no modern precedent for a midterm electorate having a higher turnout rate among eligible voters than the turnout rate in the most-recently held presidential election.
According to data from turnout expert Michael McDonald of the University of Florida, the average turnout rate for eligible voters in the 43 presidential elections held since 1856 is about 64%, while the average turnout in the 42 midterms held since that year is 49%. So the turnout is on average about 15 percentage points higher in the presidential than in the midterms, and the midterm turnout was never higher than the immediately previous presidential election. We went back to 1856 because that is the start of our modern two-party system, with the Republicans first fielding a presidential candidate that year to join the Democrats, a party that had existed in various forms prior to that year.
Of course, those who would have been an “eligible voter” was much different back then than it is now, with the franchise later expanding to previously disenfranchised groups like women, Black voters, and, in advance of the 1972 election, 18-20 year olds. Figure 1 shows the presidential and midterm turnout rates for this more modern time period.
Average turnout for presidential elections from 1972-2024 was 58%, and average midterm turnout in that timeframe was 41%, or a 17-point gap that’s very similar to, but slightly larger than, the presidential to midterm gap in the longer time period since 1856.
Note that turnout in both the last two presidentials and last two midterms have been high by recent standards—but even comparing the 2016 presidential election to the historically-high turnout 2018 midterm still showed a 10 percentage point smaller turnout in the midterm.
The bottom line is that the electorate in 2026 will be substantially smaller than 2024, it’s just a question of how much smaller…
2. The electorate should be whiter, older, and more educated.
Because midterm electorates are smaller than presidential electorates, it stands to reason that demographic groups that have historically had better turnout rates would make up a greater share of midterm electorates than presidential electorates. We see this with white voters in general, voters with a four-year college degree, and voters aged 65 and over.
Table 1 shows the makeup of the eight electorates by certain age, racial, and education groups from 2008-2022 (four midterms and four presidentials) from the Democratic data firm Catalist. Catalist’s reports on the makeup of the electorate are widely-cited and include analysis of state voter files that are not available to exit polls released on Election Night (2024 is not included here because that Catalist report is not yet out—neither is a similar analysis that we look forward to seeing later this year, Pew’s validated voter study).
While non-college voters have always made up a clear majority of the electorate, the midterm electorates have been a few points more college-educated than the presidential electorates. Pre-Trump, college and non-college voters (this includes people of all races), voted fairly similarly in both the 2012 presidential election and 2014 midterm election (the non-college group was slightly more Republican in each election). But starting in 2016, the education gap expanded greatly, so that by 2020, Joe Biden won college graduates 59%-41% in the two-party vote but lost non-college graduates 52%-48%, a 22-point gap in margin. The gap was similar in the 2022 congressional vote. Again, Catalist has not released its report for 2024 yet, but the Edison Research national exit poll for several media entities found a gap of a little more than 25 points between college graduates and non-graduates in 2024.
The 65 and over cohort has been growing over time, but their share of the midterm electorate was always at least 4 percentage points bigger than the previous presidential election during the 2008-2020 timeline. Likewise, the 18-29 group has always made up at least 3 percentage points less of the midterm electorate than the previous presidential—it likely is no coincidence that the drop off was smallest from 2016 to 2018, the only midterm in this timeframe conducted under a Republican president that also represented the best Democratic performance among these four midterms. Even though Donald Trump made some gains among the 18-29 group in 2024 compared to his previous performance with them, this is still a Democratic-leaning cohort, just like voters 65 and over remain Republican-leaning to some extent.
Finally, the electorate is usually a little bit whiter in the midterm compared to the presidential. Again, Trump made gains with nonwhite voters in 2024, but as a bloc, nonwhite voters are still markedly more Democratic leaning than white voters.
In this era of politics, the midterm having a higher share of college graduates than the presidential would seem to help Democrats, and the midterm having a smaller share of younger and nonwhite voters would seem to help Republicans. But an electorate’s demographic makeup does not necessarily tell us what the results will be: The 2018 midterm’s electorate was whiter and older than either the 2016 or 2020 electorates, but that was also the Democrats’ best election of the trio. Likewise, the college-educated share in 2022 was very similar to 2018 and perhaps even slightly larger, but that didn’t stop Republicans from winning the House majority that year.
3. The non-presidential party’s share of the House popular vote should go up.
Last week, in a piece on how House incumbents from the non-presidential party rarely lose in midterms, we noted that the presidential out-party almost always nets seats in the midterm. Democrats are hoping this trend continues, as they need to net just 3 seats to win the House next year.
In addition to typically netting seats, the non-presidential party also almost always sees its share of the total congressional vote go up in the midterm compared to what happened in the presidential.
Figure 2 and Table 2 show this dynamic, again going back to 1972. This shows the two-party House vote, and it corrects for unopposed seats in a given year by estimating the two-party vote in those seats. The data from 1972-2020 is from a past Crystal Ball contributor, the late Theodore S. Arrington of UNC-Charlotte, and the 2022 and 2024 data is from Split Ticket (they each use different methods to account for unopposed districts, but Arrington’s calculations and Split Ticket’s calculations for 2008-2020 are similar).
Read the rest of Kondik’s article for his charts, graphs and visuals and to see how he gets to his concluding sentence: “Overall, it would be a surprise if Democrats didn’t at least do better in the national House vote in 2026 than they did in 2024.”