Immediately after Donald Trump won a second presidential term in 2024 and Republicans held the House and captured the Senate, Democrats began counting the days until the mid-term elections. At long last, the mid-terms are upon us.
The mid-terms are typically bad news for a presidential administration, lacking the luster of a presidential election cycle and acting as a referendum. Going back to the times of Lincoln, the incumbent president’s party has lost House seats in 36 out of 41 elections. The incumbent president loses an average of four Senate seats in the midterms; as recently as 2014, Barack Obama lost nine senators.
But this is not a typical mid-term. Trump faces historic approval-rating lows due to an unpopular war in Iran, aggressive immigration enforcement, tariffs, high gas prices, and his co-opting of the Justice Department. It has created a dynamic that — according to polls — has Democratic Senate candidates competitive in historic GOP strongholds including South Carolina, Kansas, Texas, and even Mississippi.
Still, turning again to history, a sitting president with unified control of Congress has only lost control of both houses in five out of 41 mid-terms since the 1860s — in 1894 (Grover Cleveland); 1918 (Woodrow Wilson); 1946 (Harry S Truman); 1994 (Bill Clinton); and 2006 (George W. Bush).
Polls have always underestimated Trump, and Democratic optimism is guarded with November 3 looming. We asked Andy Crosby, an expert on political polls and surveying methodology, whether the polls can be trusted and which ones he is watching.
Q: Referencing general “congressional ballot” poll numbers such as from Marist, Focaldata, and YouGov, “generic” Democratic candidates are up 8 or 9 points. Have past polls similarly been positive for Democrats, only to turn out to be overly optimistic?
Crosby: One important thing to keep in mind regarding “generic party” polls is that we do not vote for generic party candidates for either political party. Instead, we vote for specific candidates. Within both parties, one candidate can be quite different than another, and we saw evidence of these differences in the primaries.
For example, in the U.S. Senate race in Texas, current candidate Ken Paxton is quite different than his Republican primary opponent, John Cornyn, even though no doubt the candidates agree on some issues. Not all Republican voters who supported Cornyn in the primary will vote for Paxton in the general election.
Similarly, in the U.S. Senate race in Michigan, current candidate Abdul El-Sayed is quite different than his Democratic primary opponent, Haley Stevens, even though no doubt the candidates agree on some issues. Similar to Texas, not all Democratic voters who supported Stevens in the primary in Michigan will vote for El-Sayed. As such, I would caution about using generic party polls to predict who will win the election.
All of the above said, the final generic ballot polls of 2024 predicted an extremely close race. For example, an NPR/PBS/Marist poll conducted October 31-November 2, 2024, showed that the generic Congressional ballot was a statistical dead heat: 48% of registered voters indicated they supported the Democratic candidate running for Congress, and 48% indicated they supported the Republican candidate.
Q: Is there a specific poll you are watching?
Crosby: Since different polls examine different races, I watch several polls. I specifically look for the polls that use probability-based sampling (where everyone has a chance of being selected), clearly disclose methods (for example, how many people were sampled, what mode was used, who was sampled, when were people sampled), and are members of the American Association for Public Opinion (AAPOR) Transparency Initiative. For example, Marist College has released several recent polls of different races across the country in recent weeks, and I trust their polls. I also generally try to look at polls that are as recent as possible, since people can change their minds leading up to the election.
Q: A historical assessment of congressional polls by The Economist this week found that the more partisan a state, the more the polls have underestimated the strength of the dominant party. How do you respond?
Crosby: One challenge to polls in highly partisan states is that we sometimes do not see as many polls, in particular when races are not expected to be close. This lack of polls leads to potential challenges gauging the electorate. For example, California is considered a solidly Democratic state, and Democratic candidate Xavier Becerra is heavily favored to win the election: a September Los Angeles Times/Berkeley IGS poll showed Becerra leading his challenger, Republican Steve Hilton, by a margin of 25 points (Becerra 58-Hilton 33). Perhaps not surprisingly, then, fewer polls ask about this race; only two polls have been conducted for the Governor’s race in the entire month of October. By comparison, Michigan is generally considered a swing state, and its U.S. Senate race is particularly close this year. Again, perhaps not surprisingly, there have been eight polls regarding the Senate race conducted in the month of October. When we have more polls — and particularly polls with strong methodology — that are conducted close to the election, we likely see a more accurate picture of the electorate.
Q: Is there a silent-but-large contingent of GOP voters who don’t show up in surveys, but show up on Election Day?
Crosby: Some historical pre-election polls have indeed underestimated Republican vote share. However, a task force report commissioned by the American Association for Public Opinion Research (AAPOR), which included representation by both academics and practitioners, concluded that the picture provided by polls regarding the 2024 election was largely accurate. In fact, the report concludes that during the 2024 election, state polls were their most accurate for any presidential election cycle since 1944. As AAPOR also notes, Democrats and Republicans have been underestimated approximately equally since the advent of modern polling in the 1930s. I still very much have confidence in high-quality polls.
Q: What are the polls telling us about Latino voters this cycle?
Crosby: Recent research released by Pew Research Center conducted September 28-October 4 found that Democratic Congressional candidates have a considerable advantage among Hispanic registered voters: about 51% of respondents preferred a Democratic U.S. House candidate, compared to 24% of respondents who favored a Republican (another 25% of respondents were not sure or supported another candidate). In addition, more Hispanic voters report being extremely motivated to vote this election cycle: 53% of Hispanic registered voters reported being extremely motivated to vote this year, compared to only 41% four years ago.
Q: How about the young white males who showed up for Trump in 2024?
Crosby: An overview of young voters broadly may be useful to start. A recent poll of young voters (ages 18-29) from the Harvard Institute of Politics conducted September 30-October 7 shows that young people broadly are quite concerned about the future; only a small proportion (9%) of all young respondents believe that the country is heading in the right direction. Young voters broadly also heavily favor Democrats as we approach the 2026 election. Among likely voters who say they will definitely vote, Democrats lead Republicans by a margin of 63% to 26%.
When we look at young men and Republicans more specifically, one notable finding is who exactly plans to vote in the 2026 election. Only approximately a third (31%) of young men indicated they will definitely vote in the 2026 election, compared with 42% in 2022. When we look at young Republican voters broadly, young Republicans who identify as part of the MAGA (Make America Great Again) movement, 50% indicated they will definitely vote in the 2026 election. However, only 31% of young Republicans who voted for Trump in 2024 would consider themselves part of the MAGA movement. Only 29% of young non-MAGA Republicans said they will definitely vote in the 2026 election. By contrast, 50% of young Democrats say they definitely plan to vote, and 66% of young Democrats who responded positively to the “Democratic Socialist” label said they definitely planned to vote.
