Polarization, Polling, and What to Watch For After the 2026 Midterms
Professor Yphtach Lelkes discusses the turn out for the midterm elections, trust in the polls, and what he’ll be watching once the votes are counted.
With the midterm elections less than a week away, much of the attention is focused on polls, close races, and which party will control Congress. Annenberg School for Communication Professor Yphtach Lelkes, co-director of the Polarization Research Lab, is also interested in who turns out in midterm elections and how partisan animosity shapes the way candidates campaign, and what it could mean once the votes are counted.
Ahead of Election Day, Lelkes reflects on what this year’s midterms may reveal about voters, candidates, and the state of American politics.
Midterm turnout is typically lower and skews toward the most engaged partisans. Does that mean the electorate that shows up is more affectively polarized than the country as a whole?
Those who turn up for midterms tend to be systematically different from those who only vote in presidential elections. They are more committed partisans, less likely to identify as ideologically moderate, and know more about politics. They also tend to be more educated, wealthier, and whiter than the population as a whole. But the biggest difference between midterm and presidential election voters is age: midterm voters are considerably older. Many of these characteristics are associated with higher levels of affective polarization. However, even if midterm voters are more affectively polarized, it doesn't necessarily mean we'll end up with more candidates who fail to uphold democratic norms. In fact, recent research suggests that older Americans are more likely to punish candidates who violate democratic norms.
How has polling for elections changed over the decades and how reliable are the polls?
Election polling has changed dramatically over the past decades. For much of the postwar period, polls relied on probability samples recruited through face-to-face or telephone interviews. Today, online surveys dominate the polling landscape. Some of these surveys still rely on probability sampling, but many use opt-in samples and are not transparent about their methods.
Despite these problems, polls have not become systematically less accurate. While polls performed poorly in 2016 and 2020, they were considerably more accurate during the 2018 and 2022 midterms and 2024 presidential election.
Even if polls are relatively accurate this year, they can still get the winner wrong when the elections are close. Right now, Democrats have a substantial lead in congressional polls, but many individual House and Senate races are closely contested. The lead is large enough in the House that I'd be surprised if Democrats don’t control the House, but I’m much less confident about the Senate. Control of the Senate could come down to a couple of very close races, where polling errors of just a few points could change the outcome.
Your Carnegie project explores whether partisan hostility comes less from genuine hatred and more from a calculated way of expressing political identity. What does that look like in a campaign season?
It depends on how popular the parties and their leaders are, and whether candidates want to associate themselves with their party or distance themselves from it. Right now, both parties are unpopular, and Trump’s approval ratings are underwater. That may create an incentive for candidates to downplay their own partisan identities and instead emphasize their opponents’. Rather than positive partisan identity signaling, where candidates emphasize that they're Democrats, Republicans, or aligned with President Trump, we may see more negative identity signaling, where candidates try to remind voters that their opponents belong to the other party or are associated with its unpopular leaders. In other words, rather than telling voters "I'm one of you," candidates may find it more useful to say "my opponent is one of them." Recent advertising data is consistent with this idea — Democrats have aired over 180,000 ads mentioning Trump, while Republicans have only aired about 16,000 ads mentioning him.
What will you be looking at the week after the election?
If Democrats win one or both chambers, I’ll be interested in any clue about what they plan on doing with their newfound power. Will they spend the next two years investigating Trump and his associates, or will they focus more on pushing a policy agenda ahead of 2028. Both can occur simultaneously, of course, but will one tactic dominate?
I’ll be watching what happens with the Supreme Court. While Justice Alito has indicated that he has indicated that he’s considering retirement, but said that he won’t step down if Democrats take the Senate. I’ll be interested in whether a pressure campaign will be successful in getting him to reconsider.
Most immediately, if Trump alleges fraud, how will Republican leaders and the Justice Department respond to those allegations. Will they use them to challenge election results or interfere with certifications? Will they amplify their attempts to give the federal government greater control over future elections?