To this end, we publish a set of methodological and ethical standards we apply to publicly released polls before surveys appear on our website. If a pollster fails to meet these standards, or refuses to answer our questions, we do not include their data in our work.

Today, FiftyPlusOne is permanently banning the use of any future data from The Public Sentiment Institute, and suspending the use of data from Patriot Polling until and unless they sufficiently respond to our methodological inquiries. We have discovered evidence that suggests both firms conducted surveys for political campaigns without disclosing their clients publicly. All surveys from each pollster that exist in our database at publication will be marked as partisan, as we cannot be confident they were conducted on a nonpartisan basis as the pollsters claim.

Additionally, we are permanently removing from our website all polls released by The Public Sentiment Institute, after the company admitted it fabricated vote-choice data for select respondents. This fabrication appears to have boosted the poll numbers of the political candidate that paid TPSI for polling services.

Subscribe to 50+1 today and get twice-weekly coverage of polls and our election forecasting model delivered straight to your inbox:

Subscribe now

1. Undisclosed partisan work

Late last week, FiftyPlusOne received a tip that TPSI had allegedly published campaign internals for the James Fishback campaign without disclosing that those polls were funded by a candidate. This is in explicit violation of our ethical standards for pollsters, which state, “Pollsters operating on behalf of a party, campaign, PAC, or other political client will clearly label their polls as such. Pollsters will clearly specify sponsorship on each poll.”

Because we flag for readers when polls are part of a campaign operation — and exclude pollsters who violate our public criteria from our site — we promptly went into investigation mode. In the course of that investigation, we identified a second pollster who may have engaged in similar activity.

We looked up public records on spending for the Fishback campaign, and uncovered the following facts.

1. On June 9, 2026, the Fishback campaign paid The Public Sentiment Institute $1,575, with the memo line “Polling/survey.” The Public Sentiment Institute later released two surveys of the Florida gubernatorial primary, on June 10, 2026 and June 30, 2026, neither of which includes any indication of sponsorship by any outside group.

2. On August 13, 2026, The Public Sentiment Institute released another survey of the Florida gubernatorial primary that included the statement “This survey was designed, funded and conducted by The Public Sentiment Institute. No candidate, campaign, party committee or outside organization sponsored, commissioned or paid for this research.”

3. On January 21, 2026, the Fishback campaign paid Lucca Ruggieri, CEO of Patriot Polling, $1,500, with the memo line “Polling.” On February 2, 2026, Patriot Polling released a survey of the Florida gubernatorial primary that does not include any indication of sponsorship by any outside group. To the contrary, the release includes the phrase “we have decided to conduct a head-to-head poll,” suggesting that the survey was conducted independently. The survey memo is written favorably toward Fishback.

4. On July 29, 2026, Patriot Polling released an internal survey of the Florida 2nd Congressional District general election, conducted on behalf of the Keith Gross campaign. An earlier Patriot Polling survey, released February 11, 2026, of the Florida 2nd Congressional District Republican primary election, did not disclose any sponsorship. We did not find any evidence of payments to Patriot Polling or Lucca Ruggieri in Gross’ FEC filings.

Last Friday, we reached out to Patriot Polling and The Public Sentiment Institute to ask whether any of the polls noted above were undisclosed campaign internals. We followed up with them this week but did not receive any additional information from the pollsters. We also contacted the Fishback and Gross campaigns with requests for comment and have not heard back.

Failure to disclose sponsorship doesn’t automatically mean we exclude a pollster from our database. There are a few other pollsters that have a history both of (a) answering our direct questions and (b) not including sponsorship information in their releases that we continue to aggregate. We simply treat all of their polls as partisan, even if they do not specify a partisan sponsor.

However, failing to respond to methodological inquiries renders a firm’s data unusable, since we cannot vet the pollster per our published methodological and ethical standards.

2. Manufacturing vote-choice data for individual respondents

After we finalized an original draft of this article that included just the findings of undisclosed partisanship above, The Public Sentiment Institute published a post-mortem of their Florida polls, which significantly overstated support for James Fishback relative both to other polls of the race and election results. The report concluded that their error was two-fold: First, TPSI says, they dramatically overestimated turnout among young voters. That is not a methods violation, as pollsters often make judgement calls about the expected electorate — but their second reason is.

TPSI admits that the firm changed the stated vote choice of certain respondents who favored a different candidate, current Lieutenant Governor Jay Collins, turning these respondents into supporters of James Fishback based on their belief that Fishback should have been polling higher. The report states:

Jay Collins was polling in the twenties in our data. That did not align with our read of the race, and we concluded it might reflect a data-quality problem specific to our text-message panel. Acting on that conclusion, we identified respondents who selected Collins but whose broader profile aligned with a competing coalition, and leaned those respondents toward James Fishback (emphasis ours).

This was the most consequential error in the project, and it was ours alone. … We reallocated real support away from a candidate (emphasis ours) on the basis of a suspicion that our own data contradicted. No survey methodology should permit a respondent’s stated preference to be reassigned because an analyst finds it implausible, and ours will not again.

This decision is an obvious violation of standard polling practices, which, as they themselves state, do not “permit a respondent’s stated preference to be reassigned because an analyst finds it implausible.” This decision is made all the more questionable by the undisclosed relationship between the pollster and the Fishback campaign.

FiftyPlusOne’s methodological and ethical standards include the following items:

1. All polls should have clear mechanisms in place to attempt to build a representative sample of the population or electorate. These mechanisms may vary; the most common is to weight the sample to a population benchmark, but other statistical or sampling techniques may also be used in place of or in addition to weighting to adjust the sample.

2. Pollsters will not knowingly select research tools or methods of analysis that yield misleading conclusions.

3. Pollsters will not falsify or fabricate data.

It is our opinion, as experts in aggregating and conducting polls, that switching individual-level vote-choice data in your poll to generate a result that is more in line with your “expectations” (and your preferred candidate) is tantamount to falsifying data, and that this behavior may have also been in violation of the other two standards cited here. We no longer have faith that TPSI is conducting honest research, and are permanently discontinuing use of its surveys. This specific poll will be removed from our database, as it was not conducted in a scientific manner.

These ethical issues coincidentally follow an instance of outright fabrication discovered earlier this week. A “polling” firm called Median Strategies, which appeared out of thin air two weeks ago and released “results” from a handful of “polls” across the country, admitted to the Los Angeles Times that its results were entirely fabricated as a “social experiment” designed to see how easy it would be to get the media and public to believe bogus data. It’s unclear if this was a real experiment or just the firm’s excuse after multiple organizations (including FiftyPlusOne) requested methodological information that it clearly did not have.

Responding to the Median Strategies story, the American Association for Public Opinion Research (AAPOR) released the following statement on Wednesday.

The American Association for Public Opinion Research, the leading association of public opinion and survey research professionals, strongly condemns the fabrication of data in any polling or public opinion research. Intentionally misleading the public is a serious breach of ethics and undermines legitimate polling and public opinion work.

Legitimate polls are based on responses from the public. [...]

“It is extremely disturbing that this organization produced and distributed fabricated data undermining trust in public opinion and election polling,” said Mary Losch, President of AAPOR. “Professional polling organizations are transparent, ethical, and work hard to provide critical information to the public and other stakeholders. It is important that we don’t conflate fraudulent activity like this with legitimate election polling.”

If that approach is good enough for AAPOR, it’s good enough for us. (In fact, our ethical standards are in part derived from AAPOR’s own Code of Professional Ethics and Practices.)

Responding to email outreach before this story was published, The Public Sentiment Institute produced the following comment:

At TPSI, we are committed to transparency and eager to clarify any inaccuracies or falsehoods.

Patriot Polling did not reply to multiple attempts at outreach.

Share

At FiftyPlusOne, we only aggregate real surveys that we can verify

At FiftyPlusOne, our goal is to provide an accurate representation of what public opinion is on any given issue or election, as measured by methodologically transparent, honestly conducted polls. We take this mission seriously and go to great lengths to vet all pollsters that appear on our page. We do not maintain that pollsters have to be accurate (we can compensate for pollster (in)accuracy in our models), but firms do have to be able to convince us that they are real.

These standards are why, for example, FiftyPlusOne never aggregated so-called “polls” from Median Strategies. When we contacted them to ask questions about their methodology and company, they didn’t respond, so we didn’t republish their data on our site.

The polling industry and aggregates like ours are built on mutual trust. We can’t guarantee that everybody in this space is operating in good faith, but we do our best to vet new pollsters, have ongoing relationships and communication with many firms and sponsors, and hopefully maintain trust with our audience that the work we publish is legitimate. If pollsters do not respond when we have questions about how they generate their data, we cannot verify that legitimacy. And aggregating data we cannot verify would violate the trust that readers put in us.

The Public Sentiment Institute is now permanently banned, and Patriot Polling is now indefinitely suspended, from the FiftyPlusOne database, aggregates, models, and forecasts. Any future firms founded or managed by Lester Tellez or Dustin Smith (TPSI) are likewise banned, as are surveys sponsored by organizations affiliated with these individuals, such as On Point Politics. Future firms founded or managed by Lucca Ruggieri or Roberto Jay (Patriot Polling) are subject to the same suspension as Patriot Polling.

In keeping with historical tradition established at FiveThirtyEight, previous polls conducted by Patriot Polling will remain on our site, as they contain important context for understanding our polling averages at the time of their release, but all such surveys will be reclassified as partisan. If Patriot Polling responds to our methodological inquiries and sufficiently assures us these issues will not happen again, we will consider lifting their suspension and resuming collection of their polls.

Paid subscribers to 50+1 get access to premium analysis, plus sortable tables and complete data access on our polling website. If you want to follow the 2026 cycle with the best data at your fingertips, become a paid subscriber.

Upgrade to a paid subscription today

Subscribe now

We’re just about 80 days from the 2026 midterm elections, and voters are starting to tune into the race a bit more. Much of the coverage surrounding the midterms so far has focused on each party’s primaries, but that leaves out a crucial demographic that is almost certain to decide ultimate control of Congress in November. With that in mind, this week’s Trendline will take a closer look at independent voters: where they stand ahead of November, and the issues they care most about.

In recent years, the number of Americans who identify as political independents has been growing, according to Gallup’s tracker. Since 2008, the percentage of Americans who consider themselves independent has risen by 10 percent, with a record 45 percent identifying as an independent in 2025. This trend is especially pronounced with young voters; majorities of both Gen Z and Millennials say they are independents, while among Baby Boomers and the Silent Generation, it’s closer to a third.

It may come as no surprise that voters who identify as politically independent have a dour view of the two parties. According to a poll from the Marquette Law School (Jul. 22 - 29, 2026), just 24 percent of “pure” independents (those that say they do not lean toward either Republicans or Democrats) have a favorable view of the Democratic party, and 15 percent have a favorable view of the Republican party. They’re also not so hot on the president: the same poll showed President Donald Trump’s favorability underwater among pure independents by 57 percentage points.

This frustration with political parties manifests in other ways, too. In a CNN/SSRS poll (May 7 - 31, 2026), independents were more likely than partisans to say they didn’t follow politics at all — 14 percent said so, compared to 4 percent of Democrats and Republicans. On the other end of the spectrum, 42 percent of Democrats said they “frequently seek out the latest political news,” a sentiment shared by 39 percent of Republicans and just 25 percent of independents.

CNN/SSRS also found that independents are less likely to be registered to vote, with 67 percent saying they were registered while more than 80 percent of partisans were. And like Gallup, they found independents to be younger than partisans, with 63 percent of independents under the age of 50.

When it comes to how they’re feeling about America, the picture is just as bad as how they feel about the two parties. According to YouGov/The Economist tracking data, 68 percent of registered independent voters say the economy is getting worse, higher than at any time in the last five years since July 2022, and 11 points higher than among registered voters overall. Similarly, 74 percent say that the country is on the wrong track, compared to 64 percent of registered voters overall.

This higher level of dissatisfaction among independents isn’t new, but it wasn’t always like this. During Trump’s first term (and before), independent voters generally looked similar to the nation in terms of their views of the country. But after Biden’s inauguration in 2021, independent voters started to turn more pessimistic than the overall electorate, routinely dragging behind the country on whether things are going in the right direction. After Trump’s second inauguration, they briefly improved their outlook, before dropping below the nation overall once again.

Given this sour outlook, it may come as no surprise that independents are not particularly fond of the president’s handling of his job. Trump’s approval rating hasn’t been above 30 with independents since early March, shortly after the latest conflict in Iran began. (For details on how this average was calculated, please refer to the methodology statement in this article.)

That low rating places Trump’s net approval among independents in his second term roughly 15.5 percentage points lower than his net overall approval rating, which currently sits at -24 percentage points. On average, throughout his second term, his approval rating among independents has been 15 points lower than his topline approval rating, a seemingly new phenomenon in the current political environment.

For example, in Trump’s first term, YouGov/Economist tracking data showed that among registered voters, the absolute difference in Trump’s net approval rating among independents and his overall net approval averaged around two percentage points, and wasn’t consistently higher or lower (his read was better with independents in 100 out of 207 polls, worse in 80, and the same in 27). In August 2018, before his first midterms, Trump’s approval with registered independents was hovering in the mid-40s in YouGov’s data, and never fell below 35 percent in the entire term. In the latest YouGov/Economist poll (Aug. 7 - 10, 2026), Trump’s approval rating among registered independents is 25 percent.

All this dissatisfaction is showing in the generic ballot polling among independents as well. In the last year, the number of independents who say they plan to vote for Democrats has grown by almost five points, and they now prefer a Democratic candidate by just over 15 points. (For details on how this average was calculated, see the methodology statement in this article.)

This 15-point margin, if it holds, is significant. In the 2018 midterms during Trump’s first term, exit polls estimated that independents voted for Democrats by a 12-point margin, and in 2022, they favored Democrats by two points. But there’s still a significant proportion of undecided voters here; even now, less than 100 days before the election, 28 percent of independents remain undecided.

There’s also a chance that a lot of these independent voters don’t show up to the polls at all. In a Change Research poll (Jul. 13 - 17, 2026), when asked to rate how motivated they were to vote on a 10-point scale, just 58 percent of pure independents rated their motivation at eight or above, compared with 89 percent of Democrats and 82 percent of Republicans. Twelve percent of independent voters rated their motivation to vote in November as a 0 out of 10, while just 2 percent of Democrats and 3 percent of Republicans said the same.

Independent voters may be the key to the midterms, with Democrats and Republicans nearly maxed out on their own party bases already. If Republicans can shore up their numbers with independents, they may stand a chance of stemming the blue wave that Democrats are hoping for. But given their general dissatisfaction with the state of the country, it may be a tall order.

Share

Other polling nuggets

• NIL isn’t just taking over college sports anymore: In a poll from Ipsos/The Associated Press (Jul. 21 - 26, 2026), 44 percent of parents of school-aged children who participate in sports said that their child earning money for their name, image, and likeness in the future played a role in their decision to enroll them in sports, including 18 percent who said it played a major role. But NIL considerations still aren’t the main reasons parents are enrolling their kids in sports: Around two-thirds said that earning an athletic scholarship or improving their college acceptance odds played a role. Parents are also concerned about potential injuries in K-12 sports — 39 percent of parents said they would encourage their child to play a sport other than football due to concussion concerns, while 35 percent said they wouldn’t and 25 percent weren’t sure.

• In a recent YouGov poll (Aug. 6 - 10, 2026), 59 percent of men said that it was important to them to be seen by others as manly or masculine, while 41 percent said it was not important. That number was slightly lower than the share of women who said it was important to them to be seen as feminine (63 percent). Women were also less concerned about being publicly shamed for their gender expression: 22 percent of men in the poll said they worried they would be judged or criticized for acting too feminine, while just 14 percent of women said they worried about being judged for acting too masculine.

• While Trump has spent years almost entirely reshaping the Republican Party in his image, GOP voters aren’t necessarily identifying themselves as closely with the president. In the latest Cygnal poll (Aug. 6 - 7, 2026), 54 percent of Republicans said they considered themselves more of a “traditional conservative Republican,” while only 44 percent said they identified more as a “Trump Republican.” On the Democratic side, 61 percent said they were more of a “progressive Democrat,” while just a third considered themselves more of a “traditional Democrat.”

• YouGov recently asked (Aug. 11, 2026) Americans how they pack their clothes in a suitcase. Most (58 percent) said they fold their clothes when packing, while 27 percent said they roll them. Clothes-packing experts (because apparently that’s a thing!) generally consider rolling your clothes to save more space than folding, though there are some trade-offs. But regardless of whether you fold or roll, at least you aren’t as chaotic as the 5 percent of Americans who say they just stuff their clothes into their suitcase, or the other 5 percent who say they don’t use a suitcase at all. Whatever gets you to the airport on time, I guess.

Polling averages update

All numbers are as of 1:00 PM Eastern on Aug. 14, 2026.

2026 U.S. House generic ballot

The Democratic margin in the 2026 U.S. House generic ballot rounded to 6 percentage points this week. Both Democrats and Republicans saw some improvement in their numbers, with Democrats increasing from 49 last week to 49.4 this week, and Republicans increasing from 42.5 to 43.3.

Trump’s job approval

President Trump’s approval rating among U.S. adults fell about a percentage point on net this week from where it was last week. His approval number sank about half a point from 36.3 to 36.8, while his disapproval rating grew from 59.9 to 60.4. He also hit an all time low this week of 36.1 to 60.8 on August 11.

Share

Paid subscribers to 50+1 get access to premium analysis, plus sortable tables and complete data access on our polling website. If you want to follow the 2026 cycle with the best data at your fingertips, become a paid subscriber.

Upgrade to a paid subscription today

But as friend of the site Nathaniel Rakich has frequently pointed out, while in raw dollars these numbers are eye-popping, most election finance trackers don’t adjust for inflation. This makes it hard to know whether spending in 2026 really is at an all-time high. So today, we’re looking at outside spending in every primary season since 2010 (when the Citizens United decision opened the floodgates to unlimited outside spending in federal elections), adjusted to June 2026 dollars, to see how things have changed over time. Is 2026 shaping up to be the most expensive election year on record?

To do this, we’re relying on data from OpenSecrets on outside spending in Federal primary races. OpenSecrets tracks every race where the total outside expenditures exceed $1,000; we’ve inflation-adjusted this from 2010 dollars to look at every partisan primary for House or Senate with at least $1,540 in outside spending in June 2026 dollars. (From here on out, all dollar figures should be read as inflation-adjusted to June 2026.)

Paid subscribers to 50+1 get access to premium analysis, plus sortable tables and complete data access on our polling website. If you want to follow the 2026 cycle with the best data at your fingertips, become a paid subscriber.

Upgrade to a paid subscription today

Outside actors are spending more money in more races

The first thing we noticed in the data is the dramatic increase in the number of races where outside groups have spent enough money that they show up in OpenSecrets’ data. In 2010, just 112 partisan primaries saw more than $1,540 in outside spending. While growth hasn’t always been steady, the number of primaries in which outside groups are spending money has slowly grown over time; by 2024, the number of races in which outside groups spent more than $1,540 had more than doubled to 241. This growth has been almost entirely in House primaries, where prior to 2018, typically fewer than 100 races saw significant outside spending, but ever since have had over 175 primaries with outside spending each cycle.

This growth is particularly high in Republican primaries. While outside spending has been growing among both parties, in every single cycle, the number of Republican primaries with more than $1,540 in outside spending exceeded the number of Democratic races, sometimes by more than 2-to-1.

The amounts outside groups are spending are increasing, too. If we consider the 50 most expensive House races in each cycle, for example, the average amount spent by outside groups has steadily grown, from less than $200,000 in 2010 to over $3.7 million in 2024. In 2010, just one House race saw more than $1 million in outside spending; in 2022 and 2024, only three of the 50 most expensive House primaries had less than $1 million in outside spending, and both cycles had three races where outside groups spent over $10 million.

Republicans are spending more and more in high-stakes Senate primaries

In 2014, incumbent Mississippi Republican Thad Cochran faced a Tea Party challenger, then-State Senator Chris McDaniel. McDaniel was endorsed by typical Tea Party actors, like Club for Growth and the Senate Conservatives Fund. In the previous cycle, a Tea Party nominee had knocked off an incumbent Republican in Indiana, only to go on to lose in November, so the GOP was taking no chances. In the 2012 Indiana primary, outside spending totaled just shy of $6.5 million. Spending more than doubled to over $13 million in the Mississippi race.

Cochran went on to defeat McDaniel in a runoff by just two percentage points. And since then, GOP spending in races with potentially problematic primary challengers has continued to grow. In 2016, nearly $16 million was spent in the New Hampshire primary where incumbent Kelly Ayotte faced a challenge from her right from former state senator Jim Rubens. In 2018, over $17 million was spent by outside groups in Wisconsin, where then-State Senator Leah Vukmir was locked in a tight race with Tea Party conservative Kevin Nicholson. In 2020, almost $18 million was spent in a close contest between incumbent Senator Roger Marshall and controversial former Kansas secretary of state Kris Kobach.

In each of these cases, the establishment candidate went on to win their race. And in 2022, when a number of races saw less mainstream candidates challenging the incumbent, outside GOP spending exploded. Seven Republican primary contests for U.S. Senate in 2022 saw more than $18 million in spending: Ohio, Pennsylvania, Missouri, Alabama, Wisconsin, North Carolina, and Arizona. These 7 contests are among the 8 Senate primaries with the most outside spending since Citizens United. (The other one is the Ohio Republican primary in 2024, in 5th place.) On average in these 7 races, outside groups spent $29,162,201 in the Republican primary. The most expensive primary of all was the 2022 Ohio Senate contest, a tight 3-way contest between now-Vice President JD Vance, then-State Senator Matt Dolan, and former Ohio treasurer Josh Mandel.

Republican groups also typically shell out more in, and in more, Senate races than Democrats. In every cycle since Citizens United other than 2020, there are more GOP races in the top 20 most expensive primaries than Democratic races (in 2020, there are 10 of each). (See the table at the end of the article for a full list of primaries and their inflation-adjusted outside spending.)

Subscribe now

Spending in nonpartisan primaries has also grown

In nonpartisan primaries, where all candidates run on the same ballot regardless of party, outside spending has also grown in the years since Citizens United. Because only four states use a nonpartisan primary system (and in 2010, only one state did – Washington), there are fewer of these races each cycle, but more and more are seeing outside groups spending to promote candidates.

On the Senate side, there are very few nonpartisan primaries to consider. Only three have ever topped $1 million: Alaska in 2022, Louisiana in 2014, and California in 2024. Just one of these races topped $20 million: the open primary to succeed Laphonza Butler in California in 2024, which cost outside groups nearly $23 million.

With over $60 million in outside spending, Michigan’s Democratic Senate primary now indeed appears to far and away top the list of most expensive primaries in history. Due to the premium Republicans place on their primary elections, Texas’ Republican Senate primary also will likely land near the top — once all the final reporting is made this year.

When the dust settles at the end of the primary cycle, and all the FEC reports are finalized, we’ll find out for sure if 2026 continues the trend of increased spending across the board, or if just a few high-profile races drove enormous dollars.

For now, what we can definitely say is that, even if you adjust for inflation, elections are getting more and more expensive every year.

Subscribe now

Want even more? Here’s the complete data we used for this article, for you to play with yourself.

Methodology

Primary expenditures by outside groups were gathered from data published by OpenSecrets. For partisan races, spending conducted either for or against Republicans was counted as spending in the Republican primary, and spending conducted either for or against Democrats was counted as spending in the Democratic primary. For nonpartisan races, all spending was included.

Inflation adjustments were made using the Bureau of Labor Statistics CPI Inflation Calculator. Each race was adjusted to June 2026 dollars1 based on the month in which the primary was held. Election dates were drawn from the Federal Election Commission.

The earliest dates of primaries in this dataset were in March 2010. In June 2026 dollars, $1,000 in March 2010 would be worth $1,534.49. We thus used $1,534.49 in inflation-adjusted dollars as the minimum for a race to appear in the dataset, to ensure comparable data over cycles.

Footnotes

One of those inputs is, of course, the polls — specifically, averages of polls. We have one set of computer programs that takes in the polls for every statewide and federal election in the country, adjusts them, and calculates an average estimate of support for each candidate over time, and then another program that reads in those estimates and feeds them through the forecast model.

Earlier this summer, I squirreled these polling averages away on a cloud computer server somewhere in the U.S. (probably in Ashburn, Virginia), where I left them to tick away each day while the team worked on getting our forecast model up. Now that it is, we had some time last week to put together a set of new graphs of these averages and publish them for everyone to use.

These averages are now up online at the FiftyPlusOne website. Below, I quickly run through our new estimates in each competitive U.S. Senate race.

Share

Georgia

Jon Ossoff is in the strongest position of any Democrat in the country, currently leading Republican opponent Mike Collins by 8.6 points. The incumbent has gained steadily since March, when the average opened at D+4.4. It peaked at D+12 on July 1 before settling back.

The caveat in Georgia (and most Senate contests so far) is that the polling is thin. Georgia has produced only five head-to-head polls this year, and they disagree sharply. A poll for Fox News by Beacon Research/Shaw & Co. had Ossoff ahead by 13 points in late June (which caused the July 1 spike in his support); three days later, Wick had him ahead by just 3.8. The most recent, Fabrizio Lee/Impact Research/AARP on July 16, split the difference at D+9 — and nothing has been published since.

Judging by current polls, this is unlikely to be a must-watch, uber-competitive race.

North Carolina

Another state with a big Democratic lead is North Carolina. Former governor Roy Cooper leads Michael Whatley, the current head of the Republican National Committee, by 8.0 points. A month ago, the race sat at D+8.7. Among the competitive races, this has been the most stable one (once we set aside a thin-polling artifact in January that briefly pushed the average above D+21, before correcting within a day)

The recent high-quality polling tells a consistent story. A New York Times/Siena University survey had Cooper at D+7 in late June, and Beacon/Shaw/Fox News had him at D+9 in July. More recently, Change Research had him at D+9 on August 1.

That is a strong position for a popular former governor in a state Republicans once expected to hold comfortably.

Maine

Incumbent Republican Senator Susan Collins currently trails former President of the Maine Senate Troy Jackson by 2.9 points — but this is the newest race with a small sample size. Our average begins just on July 6, and includes only five head-to-head polls conducted since.

The polling to date is all over the place. Public Policy Polling gives Jackson a 5-point margin. Z to A Research has come out on the low end, at R+2 the following day. Both of these polls, however, were conducted during the contest to replace Graham Platner on the ballot, and were sponsored by candidates or organizations supporting particular candidates in that convention. More recently, the University of New Hampshire has pegged Jackson at D+3, and Hart Research has the race at D+4. The average has drifted down from its D+5 opening as the Republican-leaning results have been incorporated.

But poll watchers, be warned: Collins has survived big deficits in the past, including in 2020 when challenger Sara Gideon led her by 8 points on Election Eve. Treat the polls as but a rough guide in the Pine Tree state.

Texas

State House Rep. James Talarico leads State Attorney General Ken Paxton by 2.2 points in what has become a genuine race, despite the state’s typical lean hard to the right. The average began dead even in November, slipped to R+1.3 in February, then climbed to D+5.4 by mid-May before easing back.

Texas is the most heavily polled race on the board, with 24 head-to-head surveys. The late-July results cluster tightly, and all show a narrow Democratic lead: Texas Southern University at D+2, ReconMR/Siena University/Texas A&M University at D+4, Wedgewood (an upstart firm) at D+2, and YouGov Blue at D+3.

Ohio

In Ohio, former Senator Sherrod Brown trailed Jon Husted by 2.5 points in December, but he now leads by 2.0 — a swing of 4.5 points.

Since June, the Democrat has seen a flurry of good polls, including one from Beacon/Shaw/Fox News at D+8 and another from Fabrizio, Lee & Associates/Impact Research/AARP at D+3. But Siena University/New York Times then had Husted ahead by 3 in late June, and Ohio has since gone quiet. There has not been a head-to-head poll in six weeks.

The current number therefore rests on polling that is now well over a month old, and should be treated with some caution.

Iowa

Ashley Hinson leads Josh Turek by 0.9 points in a race that has tightened considerably since the spring — and marks a big shift left since 2024. Turek trailed by 3 points in February and by as many as 4 in April. He briefly took a narrow lead in late June before slipping back. The race remains a tossup. The only poll released in the last 6 weeks in this contest, from Emerson College/Nexstar, shows Hinson ahead by 3 points.

Michigan

This is the closest race on the map. Failed 2024 Republican nominee Mike Rogers leads Abdul El-Sayed by 1.3 points after reversing a Democratic advantage that stood at D+3.3 as recently as July 8.

But the shift came largely from a single poll. The Glengariff Group’s July 24 survey had El-Sayed at 38.7 percent and Rogers at 49.0 percent — a roughly 10-point Republican margin, far outside anything other surveys in the race. But because of a dearth of recent data, when the poll entered the average on July 28, the line moved just over four points in a single day.

Subsequent polls were much closer: Mitchell had the race tied, SoCal had it at R+1, and EPIC-MRA had it at R+3. The average has since stabilized around R+1.

Whether the Glengariff result captured a real shift or merely reflected the firm’s house effect is the central question in Michigan. This is the fourth poll Glengariff has released in the race, and all four have shown Rogers ahead of El-Sayed by more than the average did at the time — by 6.1, 6.4, 4.9, and now 10.3 points.

Alaska

We don’t have an average for Alaska up yet because the primary has not happened. Our forecast model does use these pre-primary polls, but as a policy we do not publish general-election averages for races that aren’t set yet.

The overall picture

There are eight potentially competitive U.S. Senate races this year. Democrats need to win in six of them to gain control of the Senate — two in the seats they currently control, plus four flips for a total of 51 seats.

Doing some unofficial averaging in Alaska, Democrats currently lead in six of these eight races. The two in which they trail — Michigan and Iowa — are both essentially tied, and Michigan has no new data since Abdul El-Sayed and Mike Rogers won the Democratic and Republican primaries, respectively, last Tuesday. Their strongest positions, in Georgia and North Carolina, look like near locks for November.

Check out our new polling averages at FiftyPlusOne.news/averages.

Paid subscribers to 50+1 get access to premium analysis, plus sortable tables and complete data access on our polling website. If you want to follow the 2026 cycle with the best data at your fingertips, become a paid subscriber.

Upgrade to a paid subscription today

Subscribe now

In May, the average price of gas nationwide hit $4.48. That coincided with the highest inflation rate in three years, driven primarily by a protracted war with Iran that has disrupted supply-chain routes and sent prices soaring. Since 2022, Americans have consistently ranked inflation as the among the most important issues facing the country, and they haven’t been afraid to make their concerns known at the ballot box; in the 2024 election, high inflation and the state of the economy were some of the top reasons voters elected Donald Trump.

Affordability concerns have continued to dominate the electoral conversation this cycle, with House Democrats making it their signature campaign issue, and some congressional Republicans are worried their party isn’t doing enough to address rising costs. Those are good instincts: According to polls from YouGov/The Economist, Americans are rating inflation/prices as even more important now than in 2024, and by a significant margin. In October 2024, 24 percent of Americans said inflation was the most important issue facing the country. Those numbers have grown steadily over the past year, reaching 30 percent in May and jumping to 36 percent in the most recent poll released this week.

These rising concerns aren’t just happening in the abstract — polling has also shown that Americans have changed their spending habits in response to persistent inflation. Over half of respondents in a recent poll from Ipsos (Jun. 2 - 3, 2026) said they had cut back on non-essential spending after seeing an increase in gas prices and the cost of living, while over a third said they had postponed larger expenses and bought less-expensive brands.

Congress’s apparent inaction on these issues this year has only bred further discontent among the public. The recent rise in inflation following the war with Iran has corresponded with a dramatic decline in the public’s approval of Trump’s handling of the economy. Trump’s net approval on inflation and the cost of living is now underperforming his overall net approval rating by almost 20 percentage points, and has dropped more than 40 points since the start of his term, according to an average of polls compiled by FiftyPlusOne.news and computed by our very own G. Elliott Morris. That makes it his lowest-rated issue being tracked, and a liability not just for him but Republicans as a whole this November.

The public’s deep disapproval of Trump’s handling of inflation has led to an interesting development: For the first time in a decade, more Americans trust Democrats on economic issues than Republicans, according to polling from Ipsos/Reuters. In their most recent poll (Jul. 29 - Aug. 3, 2026), 37 percent of Americans trusted Democrats more on the economy, while 36 percent preferred Republicans. Since Trump took office, the net share of Americans who trust Democrats more than Republicans on the issue has risen by over 20 points. That’s a dramatic reversal for the GOP.

What does this all mean for November? Candidates of both parties are already hammering affordability messaging, trying to position themselves as fighters for voters’ pocketbooks. But Republicans have a much harder task ahead of them. Democrats in 2024 tried desperately to disentangle themselves from former President Joe Biden’s abysmal economic ratings and the sky-high inflation that had plagued much of his term, but they were ultimately dragged down with the ship, losing the Senate and the presidency. Republicans will be hard-pressed to distance themselves from Trump’s tanking approval on inflation and the cost of living, and voters’ increasing distrust of their ability to handle the economy.

The GOP can take some comfort in the fact that most Americans aren’t single-issue voters. Other issues, like health care, social security, and democracy, also rank highly in polls and will factor into voters’ choices. And of course, not everyone who disapproves of Trump’s handling of inflation will turn out and vote blue in November. While Democrats have seen gains in the generic ballot since the beginning of the year, it hasn’t been to the same magnitude as the plummet in Trump’s approval rating. However, as we’ve noted before, the generic ballot tends to start drifting toward the out-of-power party in the last few months of the campaign season, so there’s still time for Democrats to pick up support ahead of November.

Share

Other polling nuggets

• As ICE has fallen from the headlines, the public’s views of the Trump administration’s immigration enforcement activities have softened somewhat — though they are still not particularly positive. In an AP-NORC poll (Jul. 23 - 27, 2026), 55 percent of respondents said that the administration has gone too far in “sending federal immigration agents into U.S. cities,” down from 62 percent in February. Similarly, the share who said the administration has gone too far in “restricting legal immigration” fell from 54 to 51 percent between the two surveys, and the percentage who said Trump had gone too far in “deporting immigrants living in the U.S. illegally” fell from 52 to 47 percent.

• Sixteen percent of Americans have experienced homelessness, according to a new YouGov poll (Jul. 28 - 30, 2026). Sixty-one percent said that homelessness in the U.S. is a very serious problem, and another 29 percent said it’s a somewhat serious problem. As for what to do about it? Most think the federal government should step in. Seventy-seven percent said that the federal government should do more to address homelessness, and 70 percent said that state and local governments should do more. Just 11 percent said that the issue is beyond the government’s control, while 75 percent said that the government can do a lot to address homelessness.

• Sixty-eight percent of Americans say the war in Iran has not been worth the costs, according to new polling from Ipsos/The Chicago Council on Global Affairs (Jul. 10 - 21, 2026). This includes 91 percent of Democrats and 76 percent of independents. Republicans, on the other hand, disagree: 63 percent told the pollsters that the war was worth the costs, while just 36 percent said it was not. Overall, the percentage of Americans who think that Iran’s nuclear program is a critical threat has been falling, now standing at 49 percent. But the percentage of Republicans who think Iran’s nuclear program is a critical threat has steadily grown since 2023, from 56 to 68 percent.

• Americans aren’t going to the movies all that much anymore: 43 percent told YouGov (Aug. 6, 2026) that they haven’t been to a theater in over a year. Even so, a majority (51 percent) said that seeing a movie in the theater is a better experience than watching it at home. And younger folks were just as likely as their older counterparts to say the theater experience is better than watching at home, with 51 percent of those aged 18 to 29 saying movies are better in theaters, compared to 48 percent of those over the age of 65.

Polling averages update

All numbers are as of 1:00 PM Eastern on Aug. 6, 2026.

2026 U.S. House generic ballot

The Democratic margin in the 2026 U.S. House generic ballot grew this week, rounding to 7 points. The average number of likely voters saying they plan to vote for a Republican fell from 43.5 percent on July 31 to 42.5 percent today, a full point, while the number who say they plan to vote for a Democrat fell less than half a percent, from 49.4 to 49.0.

Trump’s job approval

President Trump’s approval rating among U.S. adults has recovered somewhat from his all time low last week. On July 31, he was under water by 25 percentage points, with 36.1 percent approval and 61 percent disapproval. Today, his approval sits at 36.8 percent, with disapproval at 59.9 percent.

Share

Paid subscribers to 50+1 get access to premium analysis, plus sortable tables and complete data access on our polling website. If you want to follow the 2026 cycle with the best data at your fingertips, become a paid subscriber.

Upgrade to a paid subscription today

How to best visualize probability and uncertainty is an undertaking that, well, involves a lot of uncertainty itself. Here at 50+1, we primarily want to make sure our graphics efficiently communicate to readers what the data can — and cannot — tell us about upcoming elections. But we also want to design graphics that are enjoyable, colorful, and work both on desktop computers and mobile devices.

After launching our new election forecast earlier this week, we thought it would be fitting to give readers a look behind the scenes at the decisions our team made about how to best visualize our forecasts for 2026.

Histogram

Visitors to our forecast page are greeted first with the following simulation distribution chart. This shows a subset of the 40,000 simulations our model makes of possible election outcomes, colored by which party wins the House (or Senate) in each simulation:

Traditional histograms in statistics are bar charts used to show a distribution of outcomes, where groups of like outcomes are assigned to a bar on the x-axis and the height of each bar reflects the frequency of occurrences in each of these bins. (For us, because we are running individual simulations, it’s important to emphasize the individual outcomes in each case.)) My starting point for visualizing statistical concepts is both with the source and the understanding of how statisticians visualize their data. I look at charts from research papers and try to simplify those down to their most important parts, writing down the questions that come to my mind when I first read one and addressing those questions in my design so that they are already answered for the next audience. From there, I take domain-specific knowledge and try to simplify it to cater to a wide audience — in a way that is approachable but still preserves the integrity of the data.

Because visualizing election simulations is a very niche concept, I saw fit to make a couple of changes when designing our histogram. First, I assign each possible House outcome one value on the x-axis instead of grouping margins into bins like D+5 - D+1 and R+1 - R+5, since the probability of each single seat outcome is of great importance. Then, in order to make it clear that the data here is a group of many individual model simulations, I show each of our simulations as their own squares — rather than totaling them and drawing a single bar for that outcome, as in the normal histogram..

I originally had this idea at FiveThirtyEight when designing the 2024 election forecast. I picked up a Galton board — one of many that were lying around — and flipped it over. When the balls fell into the shape of the bell curve I was working to create, I had this idea for the deconstructed histogram.

We also spent time deciding what the annotation layer of our histogram should be. The annotation layer introduces readers to the language of the visual. Instead of leaving the audience to parse x and y axes, the annotation explains what one visual element reads and leaves the reader with the verbiage to explore the rest of the chart.

Subscribe now

Map

The data: Odds and race ratings of each House and Senate race displayed geographically

When visualizing geographical data, there are compromises to be made — how much emphasis do we put on geography vs population? The number-one principle is to understand that land mass does not equate to population. This is why chloropleth maps are not the best solution when visualizing voting and representation.

So how do we give equal populations equal weight despite land mass? Some have solved this by arranging squares or hexagons into a suggestion of that state’s geography. That solves the “land doesn’t vote” issue, but a reader loses the geographical context they are used to.

The next question to ask is: “how do we combine both population data with geography?” The first instance I personally saw of a news outlet solving this was the French newspaper, Le Monde over 10 years ago. This is their 2024 version. The squares represent race outcomes but that is still overlaid with the geography with the U.S.

This is why our map overlays circles placed at each district’s centroid with the spatial. Of course this makes it hard to navigate New England states and densely populated areas like New York City and Los Angeles. To solve for this, we have a search for districts and candidate names, plus the interaction flow is set to zoom in on a state before a district can be selected. I want to shoutout the Economist who had the same idea for their geographic prediction model graphic and arranges the circles around densely populated areas nicely.

Other design decisions rest on mobile usability. We anchor the tooltips to the bottom of a map instead of having them float by the district on mouse hover or tap. This prevents overlap and hidden touchpoints when a tooltip is visible. It solves for limitations on mobile like screen size and a variety of different interactions like swiping (horizontal and vertical), tapping, and zooming. We also limit interactivity by showing half of the U.S. on smaller screen sizes and preventing any pinch-to-zoom or swiping by offering a button to toggle between East and West.

Subscribe now

Table

The data: everything else: odds, race ratings, demographic information, Trump margins

Tables are where we want to present a reader with everything else that is important to understand the forecast. We give users a chamber-wide overview and a district-level view of the forecast. Even through visualizations we’re telling a story, and an inverted pyramid in journalistic storytelling gives readers the top-level information first and more detailed, personalized information the further they navigate down the page.

Tables are for the “power users,” so our table is searchable by state, candidate name, and district. We give readers an initial view of what we think is the most significant information (a list of toss-up races.) But we offer preset filters for race ratings to limit overwhelming a casual reader with too many filtering and searching options.

For mobile, we switch to a card-like view like our polls table so that readers can view all of the data without having to scroll left-to-right while also preserving the position to keep it scanable.

With data so rich and complex, the visualization options and decisions are endless. With every visualization compromises have to be made in what to show and how to show it, but at 50+1 we aim for our visuals to reflect the complexity of the statistical rigor that goes into our models and analysis. Stay tuned for more forecast visualizations.

Paid subscribers to 50+1 get access to premium analysis, plus sortable tables and complete data access on our polling website. If you want to follow the 2026 cycle with the best data at your fingertips, become a paid subscriber.

Upgrade to a paid subscription today

As of 6:00 AM Eastern on Monday, August 3, 2026, the model gives Democrats an 85% chance of winning the majority of seats in the House, and a 55% chance of winning control of the Senate (51 seats or more, unless an independent wins in Nebraska, Montana, Idaho, or South Dakota). Our median forecast House popular vote margin for the Democrats is +7 points, plus or minus 5, and we forecast the party to win a median of 230 House seats (a range of 211 to 253) and 51 Senate seats (47-55).

None of this is a prediction that Democrats will win, but rather a probabilistic statement about the future if typical predictors of election results are as accurate (or inaccurate) as they have been historically. Our model generates a range of plausible outcomes for each chamber based on how often parties with similar polling leads as we’re seeing for Democrats today have gone on to win elections in the past. You can think of our forecast as an answer to the question “When Democrats have led the generic ballot by 6 points in August while Republicans have held the White House and [House/Senate] polls looked like they do now, how often have they won control of the [House/Senate]?”

This forecast is the intellectual successor to the congressional forecasting model I built at FiveThirtyEight in 2024 — which was the most accurate of any statistical forecast model last cycle and even beat the prediction markets. Our forecast combines polls, district- and state-level fundraising and past election returns, candidate-level data, and ratings from expert race-raters to produce a probability that either major party will win either chamber this November. (I say intellectual successor and not direct successor because I have had to build a new model from the ground up.)

If you would like to learn more about how the model works, you can find a methodology explainer here. (Though reader beware; we have erred on the side of verbosity.) A few highlights from the model follow.

FiftyPlusOne is aggregating polls and calculating polling averages and election forecasts for nearly every major statewide and federal election. Subscribe so you don’t miss out:

Subscribe now

1. Democrats are even better positioned than they were before the 2018 wave

We ran our forecasting model on the 2018 cycle using only the data that existed at this same point on the calendar, 92 days before Election Day, and compared it to today. It may come as a surprise to readers to learn that Democrats are even likelier in our model today to win control of the House than they were at this point in 2018, the canonical “blue wave” cycle. At this point in 2018, our model would have given Democrats a 76% chance of winning control of the chamber.

The party back then was running about 6 points ahead in the polls, as it is today, and we were also forecasting it to win 230 seats in the chamber, identical to today’s forecast.

So what changed? Mostly, the geography of competition. Compared to 2018 there are more Lean Republican seats for the Democrats to pick up — and fewer Lean Democratic seats for them to lose. Our forecast also has less uncertainty now, owing to higher polarization of elections. The combination is that in 2018, at this same point, our model saw a wide spread of outcomes: anywhere from 201 to 264 House seats. Today that spread is meaningfully tighter, at 211 to 253 Democratic seats — with far more simulations where Democrats clear the 218 seats needed for the majority.

2. A friendly map is not enough to save Republicans in the Senate

Despite the fact that most Senate seats up for grabs were won handsomely by Donald Trump in 2024, these races have turned out to be marginal in our election forecast. This is because a forecasted D+7 popular vote margin for the Democrats puts them within reach, and recruiting successes for the party have pushed them further into play.

Take the six seats most likely to decide the chamber. If you combine each state’s partisan lean — a measure of how friendly it is to Democrats/Republicans in a neutral year — with the current generic-ballot margin, four of them still lean Republican, by an average of 6 points. That’s Alaska, Iowa, Texas, and Ohio — in that order. (In North Carolina and Maine, Democratic candidates are favored with this simple formula.)

Yet in every one of these races but Maine, the actual polling has Democrats running well ahead of where the map says they should be. In Texas, for example, the lean and the environment together say the Republican candidate, Ken Paxton, should be ahead of Democratic challenger James Talarico by up six points. Yet the polls have the latter ahead by three. In Alaska, Ohio, and Iowa, polls tell pretty much the same story, with Democratic candidates polling seven to nine points to the left of their states’ partisan baseline.

The one exception in in Maine, where Susan Collins’s decades-long personal brand keeps her about ten points ahead of a state that otherwise leans sharply Democratic. Still, polls there lead toward challenger Troy Jackson by 3-4 points.

Early Senate polling is thin and tends to tighten by November, so our model doesn’t take these leads at face value — but a gap this consistent is a real signal that strong Democratic recruiting has put these seats in play.

3. Democrats sweep both chambers about half the time

One cool way our model works is that the House and Senate are predicted simultaneously, in one combined model. This innovation in forecasting (most modelers forecast the two separately) gives us the ability to calculate the “joint probability” that Democrats sweep both the House and Senate in November.

The chart below shows all 40,000 of our model’s simulated elections as of publication, lined up along the horizontal axis by the Democrats’ margin in the national House vote and colored by who controls Congress in each simulation. Because we simulate the two chambers together, we can just count the outcomes for each national popular vote: Democrats win both chambers in 53% of simulations (dark blue); they take the House but fall short in the Senate in another 33% (light blue) — the likeliest path to divided control; Republicans hold both in 12% (dark red); and the rarest result of all — a Republican House alongside a Democratic Senate — turns up just 2% of the time (salmon).

The map also gives us a good mental benchmark for what Democrats need to hit in the national popular vote to be favored to win both chambers. When Democrats win the NPV by less than about 3.5-4 points, they're underdogs to hold the House at all and unlikely to take the Senate. Between roughly 4 and 6-7 points, they are likely carry the House but Republicans keep the Senate. Once Democrats cross a threshold of about 6-7 points in the NPV, they are favored to win both chambers.

Today’s forecast — a Democratic edge of about 7 points — sits right on that upper boundary, which is why full control is at a pure toss-up in our model.

4. Our track record is very good

Lastly, as I mentioned, the models I have built to forecast congressional elections in the past have been very well calibrated — so much so that betting on them in prediction markets would have netted you money. (This is not financial advice.

This year, we built a very extensive set of backtesting programs to understand how our model would have behaved in past elections, trained only on the data that would have existed before those contests. The results are shown here:

While backtesting does not guarantee future performance, the fact that the model behaves well on out-of-sample data gives us confidence in its explanatory and predictive powers on future elections. Our model’s conclusions are roughly in line with those of other high-caliber statistical forecasters (including where I used to work).

. . .

That’s it for now, but expect more coverage of our forecast on this Substack regularly in the weeks to come. You can find the latest odds here and explore our polling averages for general elections to the House, Senate, and governorships here.

Paid subscribers to 50+1 get access to premium analysis, plus sortable tables and complete data access on our polling website. If you want to follow the 2026 cycle with the best data at your fingertips, become a paid subscriber.

Upgrade to a paid subscription today

You can read more about the state of the midterm race in our launch post here.

This article documents how we built the model(s) behind this forecast. We show what goes into it, how the model processes those inputs, and why you should trust us to provide a well-calibrated read on what public data currently says about the state of all 470 elections (435 House + 35 Senate) across the nation.

We also make these estimates available via API download for people who are interested in consuming our forecast programmatically. Feel free to get in touch with us about acquiring this data. Our business partners help fund FiftyPlusOne’s overall mission of safeguarding the collection and aggregation of all public polling in the U.S. for the foreseeable future.

This model methodology has three parts. First, I explain the inputs of the model. Then, I describe how the model transforms those inputs into one cohesive forecast of each candidate’s vote share on election day. Finally, I describe how our model takes potential error in the polls and other predictors into account.

But before that, I want to explain why we are publishing an election forecast in the first place. Then, I show the proof that it works up front.

FiftyPlusOne is aggregating polls and calculating polling averages and election forecasts for nearly every major statewide and federal election. Subscribe so you don’t miss out:

Subscribe now

1. Why forecast?

One might want to forecast an election for many reasons. Maybe you are a campaign or super PAC and want to determine, statistically, which races are the closest so you can allocate resources effectively. Maybe, as was the case for George Gallup and his contemporaries, you are academically interested in how accurate polls are. You may also think that synthesizing polling and fundamentals data could help you learn about how and why people make decisions when voting. Or perhaps you are a degenerate gambler and would like to wager against the prediction markets. Whatever the reason, the popularity of election forecasting is itself revealed demand for the tools.

However, none of these reasons are why we at FiftyPlusOne forecast election outcomes. Yes, of course, we want to accurately anticipate the result of the election — but as mentioned, our fundamental goal with this business is to be good stewards of polling data. We believe in the democratic value of public polls and their centrality to things like understanding voters’ concerns after elections. It follows that properly understanding the error that is inherent in preelection polling in particular — the limit of most of the public’s exposure to polls — is crucial so that people don’t lose confidence in surveys produced for other purposes.

In other words, we aim to produce a well-calibrated prediction of what an election might look like if public polling is “wrong” the way polls have historically erred (typically in one direction across states/contests), such that we minimize collective shock at errors that lead people to disavow polling entirely.

And if you don’t buy into that, I have a simpler explanation for you: If smart people don’t forecast elections, it won’t mean there are no election forecasts; it’ll just mean there are bad election forecasts. Further, we think forecasting is fun.

We have built what I believe to be the most comprehensive and truly Bayesian model of election outcomes available in U.S. media today. Here’s how it works.

2. Model overview and backtest record

The FiftyPlusOne method for forecasting the House and Senate draws on existing journalistic approaches to election forecasting as well as the academic literature on Bayesian models and correlated election polling errors. Our forecast “model” is in fact four models — three models of (1) what the polls say, (2) what the historical fundamentals say, and (3) what professional race handicappers say, plus (4) an “ensemble” model that combines the out-of-sample forecasts from each component model according to how much each has erred against past elections. We can then use the historical error of the ensemble model to simulate the whole election many thousands of times to work out not just who is ahead, but how wrong polls and other predictors might be, and how the race would move if they are.

Formally, our approach is called Bayesian model stacking (we shied away from the most complex version, Bayesian hierarchical stacking, for lack of development time) and is rooted in academic work showing that combining predicted probability from competing models leads to better forecasts on out-of-sample data. While some election forecasters gather time series of predictors, ignore their uncertainty, and, for example, stick them in a weighted average, our model learns the weight it should put on each component forecast based on the total uncertainty each model has about the outcome of the election — including, in the case of the polling model, the role of factors such as partisan nonresponse.

On top of this insight, we add a step that calibrates the ensemble itself to account for how forecasting errors have been shared across similar states and districts over time.

But enough about the Bayesians and their statistical tricks. I detail more about how the model works in section 3 below. What we really care about is how the model performs.

I have designed no fewer than four national election forecasting models from scratch for major media and industry clients over the last decade. (This was not by choice. Each time I have left an employer, I have had to build a new one). My presidential election model for The Economist beat the competition in the 2020 presidential race and, at ABC News/FiveThirtyEight in 2024, my House and Senate model outperformed both (a) all other statistical forecasters and (b) all the prediction markets in a series of third-party tests. The FiveThirtyEight presidential model in 2024 was also better calibrated than prediction markets in anticipating Trump’s chance of winning the popular vote (about 30 percent in our model, and about 25 percent on Polymarket and Kalshi).

I mention this not to brag, but to point out that the statistical foundations of this new approach (Bayesian methods with model ensembling) have a proven track record. But to further gauge expected accuracy, I also ran our forecast on every federal election from 2012 to 2024 in “out-of-sample” mode — that is, only factoring in the data we had at that point in time (so, in July of 2012, none of the elections from 2012 through 2026). More details on how these tests were conducted are found in section 5.

The out-of-sample forecasts according to our model are as such for both the House and Senate from 2012 to 2024:

There are some cycles you could quibble with (and a few more months of model tinkering may increase accuracy), but this record is, in precise mathematical terms, completely okay with us, especially for our purposes. No statistical forecaster, for example, predicted a Democratic House in 2024, and few were as confident in the party’s chances of holding the Senate in 2022 as we were.

Our forecasts are also well-calibrated at the seat level. On average, a candidate we said was favored to win has lost their race in just 83 out of 2,129 out-of-sample forecasts. Our historical accuracy is 96.2 percent on 1,959 House forecasts issued in 2016-2024 (when most other forecasts were active), and 95.3 percent on 170 Senate forecasts.

Excluding toss-up races (where the probability for the leading candidate is less than 65 percent), our model has never made a wrong “call” in the Senate, and has erred in just eight House races.

While we cannot be sure that future performance will exactly match these historical backtests, I am confident we have done our due diligence in accurately checking the calibration of our novel method. Additionally, the accuracy statistics of 96.2 percent (House) and 95.3 percent (Senate) and Brier score of 0.029 (House) and 0.049 (Senate) rival those of the best public forecasts in the world. (In fact, a predecessor of this model I created for FiveThirtyEight in 2024 generated the most accurate House and Senate election forecasts the outlet ever issued.)

If you want more of the nitty-gritty details of our methodology, read on. To view our 2026 election forecasts and polling averages for the House and Senate, click here.

Paid subscribers to 50+1 get access to premium analysis, plus sortable tables and complete data access on our polling website. If you want to follow the 2026 cycle with the best data at your fingertips, become a paid subscriber.

Upgrade to a paid subscription today

3. What we feed the model

In figuring out who is ahead in an election, forecasters typically consult three sources.

First, there are the polls. Polls are great predictors — generally better than the others that follow — because they tell us what voters in a specific race actually say they will do. Where there are recent, competently conducted polls, nothing beats them. The problem is that not all polls are created equal, and most races do not have any — the vast majority of the 435 House districts will never be polled publicly, not even once.

Then there are the fundamentals, predictors social scientists have long included in their models for their predictive abilities. Voters in a House or Senate seat typically vote similarly to how they did the last time around, for example, so forecasters might use the lagged result in a seat or the Democratic candidate’s margin in the last two presidential elections as a predictor in a model. Candidates who are incumbents also typically perform better, and those who don’t raise much money from individual contributors also do worse. Other predictors matter too, and are detailed in section 3.2.

Another good source of preelection forecasting alpha is expert race ratings from major handicappers — Cook Political Report, Sabato’s Crystal Ball, and Inside Elections — which classify each race from Safe Democratic to Safe Republican. These are human judgments and fallible, but experts often factor in information the other two predictor sets cannot. Handicappers talk to people who run campaigns, the candidates themselves, and are plugged into insider group chats and email chains where non-public polls abound. Even if you have polls and the fundamentals, accounting for expert ratings typically helps performance.

Finally, a word on prediction markets. Other forecasters are increasingly using prediction markets as signals in their forecasting systems. We do not, for reasons I will explain in detail a separate blog post soon. The general position, however, is summed up here: There is a poisonous logical circularity in using prediction markets to update a statistical model when the statistical model is itself being factored into the prediction market. Because of these limitations, our opinion is that if you’re going to inject prediction market data into your forecasting model, you should at least let users view a version that doesn’t rely on such data.

Our model draws on historical polling data going back to 1998, hand-entered race ratings and candidate filings dating to the same year, fundraising figures from FEC filings, and past election results, district boundaries, and demographic data from multiple sources. (Special shoutout to the folks at The Downballot for providing us some early data on presidential vote history and demographics for new congressional districts.)

3.1. The polls

With few exceptions, FiftyPlusOne uses every publicly available poll of the 2026 congressional elections. Each survey is then weighted and adjusted for several factors that might influence its results and the weight it has in our model. We consider:

• When was the poll taken? A poll from March is less predictive of how people will vote than one from October. We thus decrease the weight on older data (by about 3 percent per day early in the cycle, and 14 percent in the final two weeks), and apply a trendline adjustment to older polls to account for how the national environment may have moved since it was taken. The details of the trendline adjustment are given in section 3.1.1.

• Who conducted the poll? Some pollsters produce results that are consistently between 1 and 4 percentage points friendlier to one party over the other. We account for this with what we call a “pollster effect” or “house effect” adjustment. We measure the tendency for each polling firm to produce numbers above or below the average in each race. But, in an improvement relative to other forecasters, we fit the house effect adjustment not only at the individual race level but across races in a given cycle. This gives the model more information about how a pollster may be thumbing the scale, and is the primary reason our model was correctly bullish on Democrats to take the Senate in 2022.

• How did the pollster interview people? We also apply a “mode effect” adjustment, accounting for the tendency of polls produced via different methodologies (live phone, text-to-web, online, etc.) to produce results that are biased in the same direction. This adjustment is fit identically to the house effect adjustment, but groups surveys by sampling and interview methodology instead of by polling firm. It is generally much smaller than the house effect adjustment, with an average mode effect of less than 1 point across cycles.

• Who was surveyed? Polls of all adults, registered voters, and likely voters consistently give different answers. We estimate the systematic gaps between them and adjust polls of adults and registered voters toward the population of likely voters. Typically this adjustment favors Republicans, but in 2026 Democrats have been doing 1-2 points better in polls of likely vs. registered voters.

• Who paid for the poll? Looking at all House and Senate polls released since 1998, our model finds polls sponsored by campaigns and partisan groups overestimate support for their side by about 2 points. We shift these polls back toward the other party accordingly and give them less weight in the model (by about half).

That’s pretty much it in terms of processing polling data, with a few big flags. First, we do not use results of “synthetic” or AI-generated polls — primarily for the same reasons we do not use polls generated by statistical models (such as MRP) that distill national surveys into statewide or district-level estimates; neither are generated by processes that produce independent and normally distributed results, which our models need to capture uncertainty in polling data. And in the case of AI-generated data, we consider these to be predictions not rooted in sampling public opinion, for which we have no track record to gauge accuracy. In the past, we have likened using someone else’s predictions in our model as “using someone else’s barbecue sauce as an ingredient in your own.”

Additionally, we do not use polls that test hypothetical races, such as of a general election that includes a third candidate who has not entered a race yet. If a pollster surveys a race and presents results testing both a two-way matchup between major candidates and a multi-way matchup that includes minor-party candidates that are also running (such as a Libertarian candidate), our model uses the multi-way ballot and disregards the restrictive two-way.

Finally, we do not include polls from firms that fail to meet our methodological and ethical standards. This omits a small number of non-influential data points from our models.

3.1.1. From polls to a polling average

Because polls are subject to various sources of noise, we combine them for each race using an aggregation model. The average in a given race is built using a three-pass exponentially weighted moving average (EWMA), the steps of which are as follows:

1. Windowing, weighting, and adjusting. For any given day, we include all polls of a race that were fielded in the past 90 days. Each is weighted according to its recency (with steeper down-weighting for older polls as we get closer to Election Day) and adjusted for its population. We then apply the trendline adjustment to these corrected poll observations by adjusting each poll for movement in the generic ballot since it was conducted. If, for example, a poll is three weeks old and Democrats have gained 2 points of vote share in the generic ballot in those three weeks, we add 2 points to the Democratic candidate’s vote share in the poll (and subtract 2 from the Republican).

2. Pass 1: trendline-adjusted average. We can then calculate the support for a given candidate on day d by calculating a weighted average of all the surveys from day d-90 to d. Steps 0 and 1 are then repeated on every day of a campaign, which gives us our “first pass” of the EWMA for a given race in a given cycle.

3. Pass 2: house effects. We then adjust each poll for its pollster’s house effect. Each firm’s house effect is calculated in three steps. First, we measure the pollster’s cycle-wide lean: its average deviation from the first-pass averages across every race it polled this cycle. Because a lean measured from just a handful of polls could easily be noise, we shrink it toward zero in proportion to poll volume, keeping n/(n+5) of the measured lean for a firm with n polls — so a pollster needs about five polls before we take its lean at half strength, while a firm with 50 polls keeps over 90 percent of its measured lean.
   
   Second, we allow the house effect to vary by race. We measure how much the pollster’s lean in a specific race differs from its cycle-wide lean, and add back n/(n+20) of that difference, where n is now the firm’s poll count in that race alone. The much larger constant makes this term deliberately skeptical: A race-specific quirk measured from two polls keeps less than a tenth of its value, because house effects are mostly a property of the firm, not the firm-race pair; a pollster that runs 3 points friendly to one party tends to do so everywhere.
   
   Finally, the cycle-wide lean is itself softly anchored to a prior built from the firm’s lean in past cycles, with each older cycle counting half as much as the one after it. This historical prior is deliberately weak — capped at about a fifth of the blend in the House and a tenth in the Senate — so a firm’s past behavior informs the estimate, but does not outweigh the evidence of the current cycle.

4. Pass 3: population adjustment. Last, we align every poll to a likely-voter population frame. We estimate the gap between likely-voter polls and registered-voter and all-adult polls in the current cycle, shrink those gaps toward their historical values for each chamber, and shift non-likely voter polls accordingly. (This is the same adjustment described in section 3.1; polls of likely voters pass through untouched.)

One wrinkle: We do not show averages for a race on our website unless that race has three polls from two separate pollsters, but the version we feed into the forecast model has no such gate. This has two effects: First, you may see slight differences in our polls-based forecast for a race and the average for it. These are separate models serving separate purposes. And second, it increases the sparsity of the polling signal for the eventual model. The “polling average” from contests with fewer polls thus deserves less weight in our forecast. The next section explains how we account for that.

3.1.2. Inferring the polling average in unpolled seats (the “now-cast”)

A polling average only knows about seats that have actually been polled. But what if a race has no polls at all? For the many races nobody surveys, we use our now-cast polling model, which we published in early July, to estimate what a poll would say based on surveys in similar seats. The linked post contains a full technical description of the now-cast model, which I recap here.

The now-cast’s job is to predict what the polls would say in a seat if it were polled today. The best predictors of that outcome are (a) the generic ballot today and (b) the way the seat would have voted, relative to the nation as a whole, in the past presidential election. We calculate this for every seat, polled and unpolled, and use it as the starting point for the now-cast.

Then, we use a regression model to predict the results of actual district-level polls given that starting point. This model also considers whether an incumbent is running in a seat, and how that incumbent has performed in previous elections. We can then use the coefficients from that model — which capture how much better or worse Democrats and Republicans are doing nationwide, plus the residual for incumbents from each party, relative to baseline expectations — to predict the polling average in the remaining, unpolled seats.

For a seat with polls, the now-cast reanchors the predicted polling average to the seat’s real polling average. This anchoring is loose when there is only one poll in a seat and very tight as the polls pile up. Concretely, the anchor allows about 2 points of slack at one poll, tightening rapidly to a floor of roughly a third of a point — an allowance for residual house effects and party overperformance — by about five polls.

The now-cast makes its projections for all House and Senate seats simultaneously. It simulates the election thousands of times based on how off the now-cast has been in past elections, and how much uncertainty there is today about the polling average in each seat.

3.1.3. From now-cast to forecast

Finally, the now-cast looks ahead to Election Day. This produces a polls-only forecast that can be used to calculate win probabilities for each candidate in each seat.

The forecast adds three variables to the now-cast in each seat:

First, we add the expected national change in support for the party out of power over the remainder of the election. In a midterm, the president’s party has almost always lost ground between the summer and Election Day, so we nudge every seat toward the out-party — Democrats in 2026 — by the amount of swing we have historically seen from this point on the calendar. That change is measured across the 1998–2022 midterms, and in our model today is 3 points over the whole cycle, and 1.7 points from July 30. That means if our generic ballot average was D+5, the polls-only nowcast would expect the party to win by 6.7 points come November. This prediction has a margin of error of about 4 points.

Second, we add individual race-level movement to account for the chance something happens in a given contest that is not shared with other seats. The further out we are, the more room there is for things like primary outcomes and candidates dropping out to change the result of the election. The range of potential seat-level variation our model simulates as of late July is about 5 points.

Finally, seats with no polls carry an additional layer of error: Their margins are drawn from the now-cast’s imputation of what polls there would say if we had them. This imputation comes with a large amount of historical uncertainty — a range of about 9 points of margin in the House and 8 in the Senate, with heavy tails allowing for the occasional much-larger miss.

With those three additions, the polling model is complete: For every seat, polled or unpolled, our first component forecasting model produces a projected two-party vote on Election Day, and an uncertainty interval around it.

But polls are not the only source of information we have about a race. That’s where the fundamentals come in.

3.2. The fundamentals

The fundamentals model is a structural model of each seat’s two-party margin, trained only on the history of past congressional elections and no polls. The job of the fundamentals is to tell us what to expect in a race from the things we know before anyone is surveyed at all — things like how the seat votes historically, who is running, and whether they have raised a lot of money or have a strong electoral track record.

Our fundamentals model is a Bayesian regularized logistic regression model that takes in all regularly scheduled congressional elections from the past 20 years and tells us how much each of these fundamentals correlate with candidate performance. Here is everything the fundamentals model looks at. There are five predictors, in roughly the order they matter.

1. Seat partisan lean. We calculate the baseline partisanship of a seat by taking the Democratic candidate’s share of the vote in the last two presidential elections in a state or district, subtracting that candidate’s share of the national popular vote, and then blending the two observations, giving 80 percent of the weight to the more recent and 20 percent to the more distant. Then, we add that number back to 50 percent. This value tells us (in terms of two-party vote share) what we would expect to happen in a seat in a completely tied national election. In an era of nationalized, straight-ticket voting, this single variable does most of the work, mapping nearly one-to-one onto congressional margins.

2. Incumbency. We also include a variable measuring whether an incumbent is running and if so, for which party. Our model finds incumbents win a few points more of the vote share than their district’s lean alone would predict — though, as we will see, less than they used to. We program in the default assumption that first-term members, who have not yet built a full personal brand, get half the credit as other incumbents.

3. Candidate experience. We also measure whether a candidate has held prior elected office. State legislators, mayors, and former officeholders run measurably ahead of first-time politicians — usually by about a couple of points — and, like incumbency, this can only help the side that has it. We measure experience differently for the two chambers: In the House, where the distinction that matters most is simply whether a candidate has won anything before, experience is scored as a binary: one point for winning any prior elected office, zero for a first-time candidate. In the Senate, where the field ranges from state representatives to sitting governors, we use a graded zero-to-three scale: A candidate gets one point for winning state-legislative and local offices, two for being a member of the U.S. House or other major statewide official, and three if they are a current or former U.S. senator or governor. Experience points are only given to candidates that have won their office in an election; being appointed to serve as a senator is not the same as having faced voters. In 2026, a point of experience is worth about 1.3 points of two-party margin in the House and about 0.9 points in the Senate, so a maximally experienced Senate candidate facing a first-timer starts with a roughly 2.5point head start.

4. Fundraising. The fundamentals model also takes into account how much each candidate has raised from individual donors, which we parse from bulk FEC filings. Specifically, the model factors in the Democratic candidate’s share of the two parties’ combined receipts from individual contributors. We omit fundraising data from races where we have missing data, such as when a candidate hasn’t filed their paperwork on time. Fundraising is a stronger predictor of candidate performance in the Senate vs. the House.

5. Previous performance. Finally, for incumbents we include a variable we call the “personal vote,” which measures how much they have historically over- or under-performed their district’s partisan lean as calculated in the first step above. We build this variable by studying a candidate’s whole career and weighing recent performances more strongly. In the past, we have directly modeled these variables for editorial purposes, but decided to use a more parsimonious approach this year while we test out this new projection model.

Importantly, whereas other forecasters’ models often assume the value of things like incumbency and fundraising is fixed in the training period, ours allows coefficients to drift over time. As Lauderdale and Linzer (2013) argued in their study of fundamentals-based forecasting, the normal approach to fundamentals modeling produces predictions that are both less stable and less certain than their tidy regression coefficients make them look; a predictor that mattered a great deal in one year can fade in the next, and a model that pretends otherwise will be overconfident out of sample.

So rather than fit one fixed coefficient per variable, we let each one drift from election to election, and we carry its uncertainty forward into the forecast rather than hiding it. Each coefficient in the model is allowed to move as a random walk over time (specifically, we use a dynamic linear model with a logistic link function). The image below plots the influence each predictor has on a candidate’s share of the two-party vote:

For every seat, the fundamentals model produces an expected two-party vote share for the Democratic candidate and — just as importantly — an honest measure of how unsure it is. That uncertainty is not the hairline standard error of a regression coefficient; it’s the full predictive spread, which is wider, and which grows in exactly the competitive seats where the structural story is most contestable. Both the estimate and its uncertainty then go into the blend (Section 4), where the fundamentals are just one of three voices.

So far, all of the factors going into our model have been quantitative. Quant is great, but let’s talk about the qual:

3.3. Race ratings

While they are not “objective” indicators the way polls and fundamentals are, we have found a lot of value in incorporating expert race ratings into our prediction models. The three major national handicappers — the Cook Political Report, Sabato’s Crystal Ball, and Inside Elections — each place every federal election on a scale from Safe Democratic to Safe Republican. These ratings undoubtedly consult a lot of the same data we do (they may even shift according to forecasting models), but they also factor in data we don’t have, such as private campaign internals that handicappers get shown and we never do.

We produce a forecast from expert race ratings in a three-step process. First, we turn the labels into a predicted margin of victory for the candidate. This step is necessary to translate the language of the rater (e.g., “safe” or “likely”) into the language of our model (vote share). Using a historical dataset of race ratings going back to 1998, we score each rater’s call on a scale running from +3 (safe Democratic) through 0 (tossup) to −3 (safe Republican). We then take these numeric ratings and train a regression model to predict the realized election margin given the consensus rating across every seat a rater has ever graded. We estimate this model separately for each chamber. We then apply this model to current race ratings to translate them into vote shares.

Second, we correct for handicappers’ tendency to lag the national environment. By studying historical ratings we have found that handicappers tend to move too cautiously. In particular, in midterms, they are reliably slow to shift competitive races toward the party out of power. The chart below shows this dynamic. Pooling ratings from different time horizons over the last five midterms, a seat still rated a tossup 180 days out from the election finished, on average, about 5 points in the out-party’s favor. A seat Cook rated lean-against the out-party that early came in a dead heat (+0.3 for the out-party). So in tossup- and lean-rated seats, we nudge the margin implied by race ratings toward where the national environment currently is — as measured by the generic ballot — by an amount that would have historically shifted the ratings closer to election outcomes. This amount is largest early, when the handicappers have the most catching up to do, and fades to nothing by Election Day, as the ratings converge on the result.

Third, as with the other models, we measure how reliable this method is at predicting elections historically, and then account for uncertainty. We do this by running the ratings-only forecast on past elections, only training the ratings-to-margins model and environment-adjustment models on the data from previous cycles and applying them to the year in question. We then measure how accurate the model would have been in out-of-sample years, and use those values to simulate what could happen if the raters are wrong, in every race, while adjusting for predictable changes in the national environment.

This process gives us a quantitative distillation of a qualitative outside check on our other models. In some cases, race raters catch things the other models miss, and we want to use that information in our forecast too.

But how do these three forecasts get combined into one single estimate?

4. How we combine the three models

This is where things get pretty cool. Because we have trained three independent forecasting models, with point estimates and uncertainty intervals for every seat, we can combine these predictions based on the uncertainty we have about them. (Math-heavy paper here.)

We calculate an “ensemble” forecast for each seat by assigning a weight to each model’s prediction based on how accurate it is on out-of-sample data. The more precisely a source model has pinned down past races, the more it counts. Since polls are historically the best predictors of election results, the polls model has tighter uncertainty intervals than the other models, and tends to get higher weight.

By weighting by uncertainty, instead of some ad-hoc value assigned to each model uniformly, the model also puts more weight on predictions that are more confident. For example, the weight on the polls model is larger in races with a dozen recent polls (because the uncertainty on the polls forecast is lower), while a race with no polls leans more on fundamentals and ratings. And the balance of the weights also shifts naturally across the calendar: In the spring, when polls are sparse and uncertainty from potential change between spring and Election Day is high, polls naturally take a back seat. But as November approaches and the potential for polls to shift decreases, the polls model takes on more weight.

We do make several adjustments to the base weights derived from uncertainty intervals. First, we have found that non-polling indicators are less accurate in Senate races where no incumbent is on the ballot. Open Senate seats with no incumbent get less weight on both fundamentals and ratings. Additionally, when the polls in a heavily polled race disagree sharply with the structural estimates, we slightly increase the weight on the polls; this helped our model correctly anticipate Senate victories for Democrats in red states in 2012. Finally, we are careful not to over-shrink competitive House races toward safe ratings, because doing so has historically made predictions for surprise upsets that the polls saw coming worse. (We think this problem is related to the one that causes race raters to lag national data.)

By Election Day in a typical cycle, the House forecast is roughly half polls, 40 percent fundamentals, and about 10 percent ratings. The Senate leans harder on polls later, since Senate races are polled far more heavily. Four months out, those numbers roughly invert and the fundamentals lead in both chambers.

It is worth underscoring something, however: The weights above are expected values based on the number of polls and accuracy of the fundamentals and ratings in past cycles. Because the ensemble’s weights are based on the uncertainty of the input models, they are subject to deviate from the historical trends. If we have more polling this cycle than in the past, the model will be more polls-based.

At this point in the ensemble, we have point predictions in every seat. But the weighting of the component models has removed their uncertainty, so we need a method to add that back. And how do the predictions in each seat get added up to national forecasts?

4.1. How we account for uncertainty in the ensemble model

Warning, it’s about to get nerdy.

By far the thing we spent the most time on — and perhaps the most important factor, since it affects the headline probabilities readers consume and share — was correctly estimating the uncertainty in our predictions. Our forecast model has to correctly estimate uncertainty from two major sources: time and polling.

Temporal forecast error is rather trivial to assess. We simply take the historical point-in-time ensemble forecasts for each seat, and calculate the average absolute error by chamber. The errors of our ensemble and each component model are shown by forecast time horizon in the chart below.

The second source of error is trickier to get right. When polls miss in one state or district, they tend to miss in others. In 2016 and 2020, they understated support for Republicans nearly everywhere. In 2012 and 2022, they underestimated Democrats. So if our forecast is 3 points too Democratic in one Pennsylvania district, it is probably 3 points too Democratic in Ohio and Michigan too. A model that treats each race as an independent coin flip will badly understate how lopsided election night can be — it will tell you a 40-seat swing is essentially impossible, when the country produces one every decade or so.

4.1.1 Five levels of error

But the precise way to take into account these correlated errors is largely up to the modeler. By extensively evaluating our historical forecasts over the last 20 years, we have determined the following error structure is the best way to account for uncertainty in our model(s). In each simulation, our model adds values for the following errors to the predictions in each seat:

• a national error — how far off the whole forecast is this year;

• a chamber error, because the House and Senate can miss differently;

• regional errors, because polls can be fine in the Southwest and badly off in the Midwest;

• state-level errors, for the House model;

• a cluster error for groups of demographically similar House districts — built from each district’s age, income, education, urbanization, and racial composition — which lets the model miss in, say, heavily Hispanic seats without missing everywhere; and

• an individual race error, larger where there are few polls and where our three sources disagree with each other.

We calculate the amount of error in each group by measuring exactly how wrong the model has been in past elections, at each specific point on the calendar. The uncertainty we publish in July is the uncertainty this model has typically had in past Julys.

So how do we actually measure those five errors, and how do they make it into the simulation? This is the job of a final Bayesian model — which we call an error “meta-model” — whose training data is not polls or election results, but our own forecasting system’s past mistakes.

To build a training set, we run the full ensemble retroactively on every cycle since 2006 in strict out-of-sample mode and record its error for every seat, in both chambers, at every weekly snapshot from about 155 days out through election eve. That yields on the order of 100,000 seat-week forecast errors. With each error, we also record everything the meta-model needs to know about the prediction to train the error model: the cycle, the chamber, the seat’s census region and demographic cluster, how many days remained until the election, how many polls the seat had, how sharply our three component models disagreed with one another, and, crucially, how uncertain the ensemble itself claimed to be about that seat at the time.

The error meta-model then decomposes each error into the sum of the shared shocks listed above: one national shock per cycle (shared by both chambers), a chamber shock, a regional shock, a cluster shock for demographically similar House seats, and an individual seat error.

What the model is actually estimating is the standard deviation, or sigma, of a distribution that produces the amount of error we historically observe at each level. Each sigma — one for each level of error from the above list — follows a simple two-parameter curve over the calendar: a level on election eve, and a growth rate that scales as a square root of the days left until Election Day. From this we know (a) how much error belongs in each component of the forecast model, and (b) how much the error from each level increases as we get further from the election.

One short aside on the individual seat error: Unlike the other terms, which are additive draws from distributions of error, the seat-level error term is a direct function of the seat-level errors of the individual component models from step 3. It’s trained as a multiplier on the ensemble’s weighted-average uncertainty for a seat, with two fitted model parameters that inflate uncertainty (a) where polls are scarce and (b) where polls, fundamentals, and ratings point in different directions.

The error model is fit with a fat-tailed Student-t likelihood whose degrees of freedom equal the number of training cycles — a way of being honest that, however many individual seat errors we observe, we have only watched about ten elections actually happen.

4.1.2 Simulating the election

Once we know how much error belongs to each geographic level, we can use a technique called Monte Carlo simulation to figure out what might happen in the future if the data is as wrong as it has been in the past.

First, what is a Monte Carlo simulation? The name — a reference to the famous casino in Monaco — describes a surprisingly simple idea: When a forecasting problem is too complicated to solve with pen-and-paper math, you can instead play it out many times by varying the parameters of the underlying model and then analyzing the simulated results. This technique works because of the law of large numbers.

Suppose, for example, you wanted to know the probability of being dealt a full house in poker. You could derive that probability from combinatorics, or you could shuffle a deck, deal yourself five cards, write down what you got, and repeat the process a hundred thousand times; the share of hands that came up full houses would land very close to the true answer.

An election forecast poses the same problem at a much larger scale. We have 470 races, each with its own uncertainty, all tangled together by shared sources of polling and other error. There is no tidy formula to derive, given that uncertainty, the probability that Democrats win the House or Senate. So we use simulations instead: For each of 40,000 simulations, we take the seat-level predictions from our model and then add onto them the potential error from each source listed.

Each simulation “draws” one national shock, one shock per chamber, one per region, one per state, and one per House cluster according to the probability distribution of each that is spit out by the meta-error model. For example, if simulation No. 23,001 says the polls were 3 points too Democratic nationally, every seat in that simulation shifts together. On top of that, there’s the possibility that one party’s Senate recruits do better than its House recruits, and that the polls in the Midwest miss by a bit more on top of the national error. Add up all those errors in each simulation, apply them to the ensemble forecasts of median vote share in each seat, and you have yourself a spreadsheet of 40,000 different ways the election could go.

From here, the forecast model just needs to do some simple arithmetic to produce the numbers you see on our forecast page. We can count up, for instance, the share of simulations in which each candidate wins their race — that is their win probability. We can tally the number of seats each party holds in every simulated election, and the share of simulations in which that number reaches 218 in the House, or 50/51 in the Senate, is the party’s chance of winning the majority. Every forecast number you see on our forecast page comes from these simulations.

Two flags on this: First, the simulated shocks are drawn in such a manner that they cancel out on average at the state, region, and cluster level, so they can never change our central estimate or who is favored in a seat. And second, while the number of simulations is quite large at 40,000, it is not so large that it eliminates random variance in our predictions across different model runs. Sometimes the error can change by a few decimal places.

5. How we know it works

As mentioned in section 2, we know our method works well because we have tested it on past congressional elections.

The graphic below shows the match between our model’s predicted probability of victory for a candidate and their realized probability of victory, bucketing by probabilities in a certain range (e.g., 15-30 percent to win, 30-50 percent, etc.). This chart shows probabilistic calibration over the entire prediction time series, from 200 days before the election to midnight on Election Day.

At the current juncture, our probabilities are slightly underconfident at the seat level; candidates who we give a 75 percent chance of winning win their races on average about 82 percent of the time. But we are reassuringly rarely wrong. This chart shows the predicted vote share from our model for House and Senate races not rated as Safe Republican or Safe Democratic:

But how do we actually produce these historical predictions?

Everything we have described so far — the polling model, the fundamentals, the ratings crosswalk, the ensemble weights, and the error meta-model — has dozens of parameters that had to be estimated from historical data. That creates an obvious potential for data “leakage” (we sometimes call it “time-traveling” internally here at FiftyPlusOne): A model tuned on the very elections it is evaluated against will always look brilliant on paper, but will overstate accuracy for races it wasn’t fit to.

To guard against this problem of time-traveling, we borrow a concept from machine learning called a “train-validate-test split.” This process loops through each election to train the underlying regression models only on a subset of data from prior years, validates model parameters using the held-out data from prior years, and tests model performance on the target year. We actually do this method four times: once for each of the underlying component models, and again for the error simulator, using only out-of-sample predictions from prior years in the latter stage.

That’s a lot of information that’s frankly hard to parse even for experts, so here’s a workable example. To evaluate the model’s forecast of, say, the 2018 midterms in July 2018, we rewind the clock and hand the model only the information that existed as of July 2018 — the polls that had actually been published by that date, the race ratings as they stood on that day, the fundraising totals filed by then, and the historical predictors and results of only previous cycles, and only as of July in those previous cycles. The data for elections from 1998 to 2016 then enters a loop of train-validate model training. For each cycle, only the data for the prior cycle enters our set of component regression models, trained in the Bayesian programming language Stan (named after the guy who invented Monte Carlo methods), and then makes predictions for the given cycle.

After this loop is complete, we have out-of-sample predictions as of July for every cycle from 1998 to 2016, to understand the probable predictive power of the data in July of 2018. We then exit back to the outer prediction step for 2018. The forecast then passes the data for July 2018 through the various component models that were trained on 1998 to 2016 data ensembles them using the weights trained on 1998 to 2016 data, and simulates error using the predictive distributions of the error model trained on the out-of-sample predictions from 1998 to 2016.

We repeat this process for every week of every cycle, c, from 2006 to 2024, running the train-validate-test loop for all elections held over the past 20 years (c - 20), which gives us a time series of out-of-sample forecasts as we would have issued them on those days of the campaign. When the 2018 version of the model makes its predictions, it is exactly as ignorant of 2018, 2020, and 2022 as the real 2018 version would have been. There’s no time-traveling to future years.

(In the interest of full transparency: There are two small, deliberate exceptions, where a fixed table of weakly informative priors is shared across test years for computational reasons. Neither contains outcome data, and neither meaningfully moves the results. It just helps our backtest models fit faster, so it takes a day instead of weeks to train and test a given model spec.)

Across every regularly scheduled House and Senate election from 2006 through 2024, this process generates roughly 100,000 individual seat-week forecasts, each paired with the result that actually happened. That archive of predictions is the raw material for the claims we make of accuracy and calibration in section 2 of this article, and it doubles as the training data for the error meta-model described in section 4.1.

One last question a skeptical reader might ask: Couldn’t we have simply tinkered with the model until the backtest looked good? To a degree, any modeler shapes their model with one eye on historical performance (what we call “researcher degrees of freedom”).

This is a fair objection, and only true out-of-sample prediction — on future elections — can prove our model’s forecasting ability is as good as our backtesting suggests. All we can do now is promise that our computer code is coded correctly, and we hope you trust us that we are trying to produce the most accurate read of pre-election polling and other data possible, not simply take a victory lap and promise unparalleled accuracy. Remember, FiftyPlusOne’s overall mission is to provide a well-calibrated translation of polling data to probabilistic forecasts to help the public consume such polls in a smarter way, not to sell you something or make a million dollars betting on elections.

For transparency, we have also published our model’s raw backtests for public scrutiny on our GitHub page.

6. Odds and ends

A couple of small details are worth explaining to the most technical readers in our audience.

Our model makes predictions for the Democratic and Republican candidates’ share of all votes cast for one of the two major parties in a seat. This is an oversimplification of the way elections happen in reality that helps the model fit faster and predict better; it doesn’t really matter for control of a Senate seat that’s 60 percent to 35 percent in favor of Republicans whether the Libertarian is at 4 or 5 percent.

Sometimes, however, there are more than two major candidates in a race — or two major candidates from the same party. Here is how our model deals with those:

• Ranked-choice voting (Maine, Alaska). We use the final round of ranked-choice polling, comparing the leading Democrat to the leading Republican — which is how the votes actually resolve, since minor candidates are eliminated and their ballots transfer to the front-runners rather than being summed. For Maine’s Senate race, with one candidate per party, this is exact. Alaska is our one real approximation: We get the party that wins the seat right, but if two Republicans reach the final round, we do not sort out which of them wins.

• Top-two primaries (California, Washington). The top two primary finishers advance regardless of party, and eight 2026 races so far are between two candidates of the same party — seven California Democrat-vs.-Democrat contests and one California Republican-vs.-Republican (Washington state has not yet held its primary). Party control of those seats is certain; the question is which candidate wins. We model that separately, from incumbency, fundraising, and field size, trained on past same-party contests in those states.

• Louisiana’s jungle primary. This year, every Louisiana House race will put all candidates of all parties on the November ballot, with a December runoff if nobody clears 50 percent. We simulate both rounds, including runoffs between two candidates of the same party.

• Primaries that haven’t happened yet. When a party has not picked a nominee, we do not split its vote among the contenders — that would make each look like a loser even in a safe seat. Instead, each simulation picks one potential nominee, weighted by incumbency and fundraising, and then simulates the general election with that candidate. We then pool all of these possible outcomes to forecast the chance that “a Democrat” or “a Republican” wins that seat.

• Strong independents. In Montana, independent Seth Bodnar is running alongside Democrat Alani Bankhead against Republican Kurt Alme, and polling far better. Because our pipeline builds every race as Democrat versus Republican, an independent is invisible to it — which meant we were forecasting the wrong contest. We now run that seat as a mixture of two scenarios, one where the Democrat is the effective challenger and one where the independent is. How likely each is, is a judgment call we make and disclose rather than a measurement.

• No major party challenger. When a race is between an independent and a major-party candidate, our fundamentals model treats the independent as belonging to the other party. This is the case in Nebraska’s Senate race in 2024 and 2026, for example, where Democrats decided not to run a candidate and the election ended up being between independent Dan Osborn and incumbent Republican Pete Ricketts.

And that’s pretty much it! If you have questions, you can reach out to us at data@fiftyplusone.news.

Frequently Asked Questions

• Is this a prediction? No. It is a forecast, a probability distribution over outcomes. If our 70 percent favorites lost every time, the model would be broken; if they won every time, it would also be broken. Empirically, candidates we give a 70 percent chance of winning win about 70 percent of the time.

• Why does the forecast disagree with your own polling average? Because the average is only one of three inputs, and early in the cycle it is not even the heaviest one. A race can have one poll showing a tie while the fundamentals and the handicappers both say it is safe. The forecast will say it is probably safe, and history says that is the right call in July.

• Democrats keep overperforming in special elections. Why doesn’t the model show it? In our testing, we found that including special elections as a predictor of the national political environment worked until 2018, then fell apart. We think this is because the group of voters turning out in special elections is now systematically different from the group turning out in November.

• Does it account for redistricting? Yes. When a seat is redrawn, we recalculate historical election results in the new seats, and we discount the incumbent’s personal vote advantage. Historically, when an incumbent has been drawn into a new seat, they have retained about 70 percent of their incumbency advantage and personal vote. Seats that were redistricted also have slightly more uncertainty than other seats.

• The model says 3 percent for something that seems impossible. Is that a bug? Probably not. Our model prices in the chance of systematic changes in the political environment that otherwise seem impossible, such as 9/11 disrupting the 2002 midterm cycle. After 2020, when generic ballot polls were historically off, our model also simulates fat-tailed polling error to account for systematic, unforeseen partisan non-response bias.

• Why did the number move when there was no news or new data? Usually this happens when an old survey becomes so old that the polling average in a given race drops it from the training data.

• Can I get the underlying data? Yes — see our downloads page or get in touch with us.

Corrections and model changelog

Corrections. While we do not anticipate making any errors, sometimes they happen. We will document any big corrections here.

• August 7: The 2020 presidential numbers we used for Florida’s new congressional districts were wrong. All 28 of them were about 4.8 points too Republican. The corrected figures were pushed this morning, and the model has now been re-run with them.
  The effect of this error was small, because the 2020 past elections numbers only get 15% of the weight in each district’s partisan lean (the 2024 results, which were already correct, get the other 85%). Florida’s expected Democratic seats went from 6.9 to 7.2, and the chance of a Democratic House majority nudged from 89% to 90%. All seat-level probabilistic differences were under 3 percentage points except in FL-22, which moved from tilt Republican to a genuine toss-up (45% → 51% chance for the Democrat). Nothing outside Florida moved at all, and the national projected win probability and seat total for either party was unaffected. We regret the error.
  

• August 3: An error in our 2024 historical presidential election results data in California’s new congressional districts was detected and corrected on August 2. While the error was caught before publication, the model was not rerun to reflect this change publicly. This impacted most redrawn seats by less than a point, but moved the Democratic candidate’s chance of winning CA-22 up from 72% to 88%. We apologize for the oversight.

Model changelog. While we do not anticipate having to fix anything with the model, sometimes it’s necessary to make small tweaks to our data input or processing pipelines as we watch how the model responds to new data. In rare cases, we notice a problem with the underlying statistical models and have to fix them. We will document any changes to our model here.

• August 3, 2026 - Forecast published

• July 20, 2026 - Model code frozen, watching production time series for two weeks to verify forecast moves as expected with new data.

All eyes are turning to the 2026 midterms, but while November is still months away, some of the most important battles of this year’s election cycle are already underway in the primaries. These intraparty contests tell us not just which candidates will appear on the ballot this fall, but which ideological factions, political figures, and interest groups are influencing the future direction of the Democratic and Republican parties.

Through June 30, nearly 2,000 people have appeared on the ballot this year in Democratic and Republican primaries for Senate, House and governor across 31 states. Carrying on the tradition of FiveThirtyEight’s endorsement-tracking primary project that began in 2018 and ran through the 2024 election cycle, we are compiling primary race endorsements from several key endorsers — President Donald Trump, progressive leaders and organizations, party committees, women’s groups, and Israel-focused advocacy groups — to examine what types of candidates each party is actually nominating, and what that might tell us about where the parties are headed.

So far, our data tells us that both parties are being pulled by several highly energized factions. As Democrats have duked it out over Israel policy and generational change, progressives are winning primaries in numbers rivaled only by their banner year in 2018, challenging moderate and longstanding liberal members alike. Among Republicans, Trump’s backing has become his party’s gold standard — though his endorsement strategy helps explain his nearly bulletproof record. And women’s groups on both sides of the aisle are seeing success, but partisan gaps remain large, as female candidates continue to consolidate gains in the Democratic Party, while Republican women struggle to keep pace.

The table below summarizes primary win rates among all candidates endorsed by each of the groups and individuals we are tracking, through the primaries held on June 30.

Of course, there are still hundreds of primary contests yet to come. We’ll continue tracking every endorsement and every outcome to check in on how each group does at the end of the primary cycle, so stay tuned.

Paid subscribers to 50+1 get access to premium analysis, plus sortable tables and complete data access on our polling website. If you want to follow the 2026 cycle with the best data at your fingertips, become a paid subscriber.

Upgrade to a paid subscription today

Progressive candidates are on a heater

There’s no question that it’s been a good year for progressive candidates. Well-funded and little-known challengers alike have toppled Democratic incumbents or bested competitive primary fields across the country, sparking comparisons to the 2018 progressive wave that ushered “The Squad” into Congress. Across most progressive organizations and politicians we track, around two-thirds of their non-incumbent endorsees have prevailed in primaries so far this year. Of course, there was some overlap in who each group endorsed, but all in all, 51 non-incumbent candidates have been endorsed by at least one progressive group we tracked, and 28 of them won or advanced to the general election, for an overall success rate of 55 percent.

These numbers are roughly in line with the success rates of progressive endorsees in recent cycles going back to at least 2018. More notable, though, is the number of progressive candidates that are winning primaries. Every endorser we track has already seen a higher number of its non-incumbent endorsees win just halfway through the 2026 primary season than the total number of candidates they endorsed in 2024, and many are right on the heels of their total numbers from the last midterm cycle in 2022. That means more progressives will appear on general election ballots in November and, most likely, in Congress next year.

One major story surrounding the progressive movement this year has been the number of aggressive challenges to longstanding Democratic incumbents, including not only moderates but older progressive-leaning members too. Three Democratic incumbents have lost their primaries to progressive challengers this year: Adriano Espaillat and Dan Goldman in New York, and Diana DeGette in Colorado. Notably, all three are members of the Congressional Progressive Caucus; instead of attacking them broadly from their ideological left, their challengers focused specifically on each incumbents’ support for Israel — we’ll talk more about those dynamics later — as well as Espaillat and DeGette’s ages. Another CPC member, North Carolina Rep. Valerie Foushee, narrowly fended off a similar progressive challenge by just one point, and almost certainly benefitted from an early primary that took place in March, before the momentum behind progressive candidates really began to build.

California also saw a wave of generational challengers this year, and some of the most successful were progressives: Three incumbents in California — Jimmy Gomez, Doris Matsui, and Mike Thompson — each advanced through their nonpartisan top-two primaries with less than an outright majority of the vote and will still have to go head-to-head with notable progressive challengers in the November general election.

Progressives also saw their share of high-profile losses, but even some of those made a splash and helped shape the party narrative in safe-blue seats. In Colorado’s Senate primary, for example, Julie Gonzales received less institutional support from progressive groups in her longshot bid against Sen. John Hickenlooper, but overperformed by earning 47 percent of the vote despite being outspent more than 10 to 1. And in a slew of crowded open-seat races in Illinois, multiple progressive candidates helped define races and draw big spending, but also frequently split the vote, yielding some wins for more moderate candidates with only a plurality of the vote.

Most of progressives’ activity took place in deep-blue territory, where their candidates tend to have the best chance at winning a general election: Around 80 percent of progressive-endorsed candidates on our list were running in districts rated as “solid Democratic” by the Cook Political Report. However, progressive groups did go head-to-head with the party campaign apparatus to notch a couple key wins on moderate territory.

The DCCC faced intense intraparty criticism for putting money behind Jasmeet Bains in California and David Baldacci in Maine, who both lost to progressive opponents. These swing-seat faceoffs stood alone in our dataset by pitting notable progressive candidates directly against the DCCC, which endorsed exclusively in “tossup” and “lean” seats. Bains was the only DCCC endorsee to date who’s faced a progressive on our list, Randy Villegas, while Baldacci also faced two more progressive candidates, though not ones endorsed by organizations we track. (Elsewhere, all of the committee’s other endorsed candidates won in primaries through June 30, including 18 incumbents and 9 who are vying for open or Republican-held seats.)

Looking ahead to the second half of the primary season, one of the biggest progressive vs. establishment battles of the season will play out in Michigan’s Senate primary, where progressive favorite Abdul El-Sayed faces Rep. Haley Stevens, who’s earned a rare primary endorsement from Majority Leader Chuck Schumer. (His only other endorsee was Gov. Janet Mills in Maine.) Meanwhile, a bounty of House Democrats remain at serious risk of losing to notable progressive challengers.

FiftyPlusOne is tracking every endorsement through the fall — subscribe so you don't miss how these races play out.

Subscribe now

Trump’s endorsement is the GOP golden ticket

As we’ve found in the past, Trump has endorsed more primary candidates in each election cycle since his first term as president. As he continues to tout his reputation as a GOP kingmaker, his activity this year continues that trend. Altogether, the president has already endorsed 178 Republicans in primaries held through June 30 — almost as many as the total number he endorsed in 2022 and 2024. Of these, 173 endorsees have won, representing a 97 percent success rate.

As in past cycles, that number is arguably inflated by who Trump chooses to endorse: Most of his picks are incumbents (72 percent of his endorsees), who tend to win primaries at extremely high rates regardless of endorsements, or candidates with no primary opponent at all (41 percent).

Among the 128 incumbents he endorsed, only one lost: Texas Rep. Tony Gonzales, who was forced into a runoff but dropped out amid an affair scandal, handing the nomination to Brandon Herrera. Trump endorsed Gonzales before the scandal broke, but seemed to informally back off the endorsement shortly ahead of the election. (After Gonzales dropped out, he endorsed Herrera ahead of the general election.)

Even if we narrow the list to non-incumbents running in a contested primary, Trump’s record drops to a still-strong 41 out of 45 wins, or 91 percent. In two of these races, Trump technically neutralized his losses by endorsing twice: After his pick in Oklahoma’s 1st District advanced to a runoff, the candidate’s campaign collapsed amid (another) affair scandal; Trump then switched his endorsement ahead of the runoff. He also hedged his bets by endorsing both runoff candidates in the South Carolina governor’s race. Interestingly, Trump’s only two other primary endorsement losses also came in gubernatorial races: Iowa and Georgia.

But in Congress, Trump’s endorsement reigned supreme. Campaign committees traditionally play a leading role in recruiting congressional candidates and consolidating support around an eventual nominee, but a look at how the president’s endorsements map onto those made by the committees points to Trump’s dominance there.

The National Republican Congressional Committee has endorsed far fewer House candidates than Trump — just 23 — and boasts a 100 percent success rate. Notably, the committee didn’t endorse against Trump in any instance, and all but three of its endorsees also had Trump’s backing. (Two of those, Reps. David Valadao and Brian Fitzpatrick, were moderate incumbents in uncontested primary races. The other, Laurie Buckhout, won a contested five-way race that saw candidates jockeying to align themselves with Trump.) In one indication of how closely the committee has aligned itself with Trump, earlier this year it rebranded its longstanding “Young Guns” candidate development program, which supports GOP candidates in key battleground districts, as “MAGA Majority,” promising to provide support to candidates who “embody the Trump agenda.”

On the Senate side, the National Republican Senatorial Committee did take a couple swings against Trump, but lost in both cases. NRSC leadership formally endorsed eight candidates — six in open races and two incumbents. The six non-incumbents were also endorsed by Trump and won, while the two incumbents, Sens. Bill Cassidy in Louisiana and John Cornyn in Texas, lost to Trump-backed challengers Julia Letlow and Ken Paxton, respectively.

The circumstances of both Cassidy’s and Cornyn’s losses demonstrate how Republicans have increasingly looked to Trump, not party committees or congressional leaders, for public signals of who the party should rally behind. In both cases, the NRSC endorsed its incumbents in mid-2025, but that did little to head off challengers: In Louisiana, Trump publicly encouraged Letlow to enter the race in January 2026, and in Texas, party leaders reportedly worked for months to convince Trump to back Cornyn, before being snubbed by a late endorsement that seemingly sunk the incumbent’s chances.

Subscribe now

Women continue to overperform in Democratic primaries

As we’ve done in cycles past, we took a look at how female candidates are doing in primaries, tracking endorsements by women’s groups and digging into what kinds of races and seats women are running and winning in.

Endorsements by women’s groups — who recruit, train, and support female candidates — were once again a reasonably strong predictor of success for endorsees in both parties. On the Democratic side, EMILYs List has endorsed 34 candidates in primaries held so far this cycle, and 79 percent of them (27 candidates) won their primaries, much higher than the 47 percent win rate for Democratic women overall.

Republican women’s groups did even better. Candidates backed by Winning for Women won or advanced in 92 percent of their races (22 of 24), and Maggie’s List-endorsed candidates won 83 percent of theirs (24 of 29). That’s more than double the 35 percent baseline win rate for all women running in Republican primaries so far this year.

Looking specifically at open primaries — where endorsements tend to have the most value as candidates seek to stand out from the field — EMILYs list endorsees have seen somewhat lower success rates than in recent years, while GOP endorsees have rebounded from a poor showing in 2024. Of course that could change, as many contests remain and these groups endorse selectively. In races so far, EMILYs List has endorsed 22 candidates in open primaries, and won 16 of those (73 percent), Winning for Women endorsees have won six of eight open primaries (75 percent), and Maggie’s List endorsees have won nine of 14 (64 percent).

That endorsement math is one thing. The candidate math is another. And on the candidate supply side, the GOP is still trailing Democrats. So far, 321 out of 983 Democratic primary candidates have been women (33 percent), compared with 168 out of 842 Republican primary candidates (20 percent). Democratic women are even better represented among primary winners, making up 44 percent of their party’s general election nominees so far, compared to just 18 percent of Republican nominees.

That share is fairly stable for both parties compared to the last several congressional election cycles, as shown in the table below, with one exception: Women are just 27 percent of Democrats’ nominees for Senate seats thus far.

Not only are more Democratic women winning primaries; they are also winning in districts where their chances of winning the general election are better. Notably, 54 percent of Democratic nominees in races for congressional tossup seats are women, compared with just 15 percent for Republicans, which is just two candidates. Both of those Republican women are incumbents (Rep. Mariannette Miller-Meeks in Iowa’s 1st District, and Sen. Susan Collins in Maine), while Democrats have so far nominated six non-incumbent women to challenge sitting Republicans in tossup races.

There’s a big partisan gap in women running in safe seats as well. Women are 44 percent of Democratic nominees in solid-blue congressional seats, but just 14 percent of Republican nominees in solid-red seats. That pattern isn’t new: It reflects a much older pattern in how parties recruit women. Historically, women have been tapped to run as “sacrificial lambs” in less winnable races, while the most-winnable seats go to men. Democratic women have broken from that mold in recent cycles, while Republican women still seem stuck in it.

A lack of coordination and broader institutional support from the party could be another reason that Republican women’s success lags behind Democratic women’s. For instance, in Texas’s 34th District, both GOP women’s groups we’re watching backed former Rep. Mayra Flores, who Trump had endorsed in her earlier bid; this time, he endorsed against her, and she lost. Women’s groups also clashed with Trump in two other Texas districts: the 38th, where Winning for Women and Maggie’s List backed different women, and the 10th. Trump’s pick won in these contests, too. This dynamic was present in 2024 as well, when Trump endorsed against notable female candidates in primaries for North Carolina’s 13th District, Kansas’s 3rd District, and Indiana’s race for governor.

On the Democratic side, EMILYs List has also had some endorsement clashes. While they’ve only endorsed against the DCCC once so far in this cycle, they’ve endorsed against at least one progressive group in six races. They’ve also backed the same candidate as a progressive group in four races, and the same candidate as the party committee in 12 contests.

Pro-Israel groups are playing both offense and defense this year

One of the biggest dividing lines among Democrats in this year’s primary cycle this year has been over support for Israel. Pro-Israel groups like AIPAC (and its super PAC, the United Democracy Project), Democratic Majority for Israel, and J Street have each been major players in the primaries so far.

Pro-Israel groups in the last primary cycle made big moves to unseat progressive incumbents, but they’ve found that script flipped on them in at least a few races this year. Much of these groups’ activity has focused on shoring up existing allies in Congress: Each group has endorsed at least 50 incumbents so far, and saw at least a 95 percent success rate. That said, as we mentioned in our discussion of progressive endorsements, pro-Israel groups were on the losing side of some high-profile upsets, with several incumbents that they backed falling to progressive challengers who hammered them for their pro-Israel stances.

Beyond this, all three groups have seen mixed success in backing non-incumbent candidates. But they’ve also taken steps that have made it difficult to draw a clear and cohesive conclusion about the success of their endorsees this year.

That’s perhaps most true of AIPAC, one of the biggest spenders in the entire primary cycle to date and a leading force in opposing progressive candidates. According to our endorsement-tracking system — which counts AIPAC endorsements that are posted directly on its website or direct spending by UDP in support of a candidate — the group has only endorsed two non-incumbent Democrats so far this cycle: Adrian Boafo, who won in Maryland, and Melissa Conyears-Ervin, who lost in Illinois. But AIPAC and UDP support and oppose candidates in lots of other ways, including spending money against a candidate without explicitly endorsing another.

Back in February, for example, the group started off the cycle with a high-profile loss: the special election primary for New Jersey’s 11th District, where it spent over $2 million to oppose Tom Malinowski, largely through ads completely unrelated to his stance on Israel. That strategy is widely seen to have backfired and helped push a more progressive candidate across the finish line.

In other cases, AIPAC’s activity isn’t captured in our dataset because the organization and its allies often funnel their money through shell PACS that can receive funding from other sources besides AIPAC, making it difficult to track.

That was the case for multiple congressional races in Illinois, where AIPAC funneled money into multiple shell PACs with varying degrees of success. In Illinois’s 9th District, the AIPAC-affiliated Elect Chicago Women and Chicago Progressive Partnership PACs each spent millions to unsuccessfully support Laura Fine and oppose more progressive candidates Daniel Biss and Kat Abughazaleh. They also spent a small sum supporting another progressive, Bushra Amiwala, in the same race, in an attempt to split the progressive, pro-Palestinian vote. In the 8th and 2nd districts, Elect Chicago Women and another shell PAC, Affordable Chicago Now, successfully spent huge sums to push Melissa Bean and Donna Miller across the finish line, respectively. But their money couldn’t quite elevate their pick, Melissa Conyears-Ervin, above a crowded field in the 7th District.

AIPAC has also endorsed 66 incumbent Republicans in primaries so far this year, and all but four of them won. The group only backed one GOP challenger, Ed Gallrein in Kentucky’s 4th District, contributing to one of the most expensive primaries to date: the effort to unseat maverick Rep. Thomas Massie, who notably earned the ire of President Trump.

But AIPAC isn’t the only pro-Israel organization playing an active role in primaries this cycle. DMFI has been very successful, with seven of their 10 non-incument endorsees winning in contested races. A smaller organization focused solely on Democratic candidates, DMFI has often overlapped with AIPAC since its founding in 2019, but usually hasn’t played the same role of political boogeyman in primaries, even though it too is a major participant in the shell PAC game.

The third group we tracked, J Street, often breaks from the others. A more liberal pro-Israel group that supports a two-state solution and an end to war in the region, J Street has cast a broader net with its endorsements rather than going all-in on select candidates. In the first half of the year, J Street endorsed multiple candidates in the same race eight times. That’s not necessarily a new strategy for the group, though, and it seems to have worked out pretty well for them this year, even if it muddles their overall success rate. While half of their non-incumbent endorsees won in contested primaries, one of their preferred candidates won all but five of the 22 races of this type that they endorsed in.

J Street-endorsed candidates also performed well when directly up against AIPAC- or DMFI-backed candidates. While several candidates won with both J Street and another group both backing them, candidates endorsed solely by J Street beat out an AIPAC or DMFI pick in four contests. The group’s only loss in these head-to-heads was in California’s 32nd District, where J Street endorsed against incumbent Rep. Brad Sherman, while AIPAC and DMFI supported the congressman.

In the latter half of this primary season, pro-Israel groups will face big tests, particularly in Michigan’s Democratic Senate primary where AIPAC is spending heavily to boost Rep. Haley Stevens over Abdul El-Sayed, and in Missouri’s 1st District, where pro-Israel groups played a major role in ousting former Rep. Cori Bush in 2024, and are now opposing her comeback campaign.

We’ll be tracking all of these storylines throughout the rest of the primary season, and we’ll be back in the fall with more comprehensive rundowns for each. We’ll also make our full dataset available for download and reuse at that time.

In the meantime, if you’re interested in accessing our partial data through the June 30 primaries, reach out to us here.

Want the post-election primary rundown as soon as it's out? Subscribe to FiftyPlusOne and get our primary tracking, data, and analysis delivered straight to your inbox.

Paid subscribers to 50+1 get access to premium analysis, plus sortable tables and complete data access on our polling website. If you want to follow the 2026 cycle with the best data at your fingertips, become a paid subscriber.

Upgrade to a paid subscription today

CORRECTION: An earlier version of this article inadvertently classified Julia Letlow as a candidate running in an open primary, but she ran against incumbent Sen. Bill Cassidy. The table “Women’s group endorsees tend to do well in primaries” and corresponding text has been updated to reflect this.

Methodology

Endorsement data for this article was pulled from the official websites of the Congressional Progressive Caucus, Justice Democrats, the Working Families Party, the Sunrise Movement, Indivisible, Winning for Women, Maggie’s List, EMILYs List, the American Israel Public Affairs Committee, Democratic Majority for Israel, J Street, the National Republican Senatorial Committee, the National Republican Campaign Committee, and the Democratic Congressional Campaign Committee. We also sourced endorsements from Donald Trump on Truth Social, and from Chuck Schumer, Alexandria Ocasio-Cortez, Bernie Sanders, and Elizabeth Warren on X, Instagram and from news reports. (The Democratic Senatorial Campaign Committee has not formally endorsed any candidates in 2026.)

For this project, we considered an endorsement to be any statement made by Election Day in a race supporting a candidate’s campaign and/or urging people to vote for them. We included endorsements made after an initial primary but before a runoff election with a note.

Additionally, we considered spending in support of a candidate by the United Democracy Project as an endorsement of that candidate by AIPAC, since the UDP explicitly bills itself as the super PAC of AIPAC, and AIPAC rarely, if ever, formally endorses non-incumbent primary candidates.

Candidate information and certified election results were sourced from official state election websites. We did not include races that went to a primary runoff that has not yet been decided. When calculating primary success rates, we generally included candidates who won their partisan primary or advanced to the general election. In California, the top two primary finishers, regardless of party, advance to the general election. Race ratings for individual seats were sourced from the Cook Political Report. Gender for candidates was sourced from the Center for American Women in Politics at Rutgers University.

Historical endorsement data was sourced from FiveThirtyEight’s public datasets.

Upgrade to support 50+1

There’s been a lot of talk in recent weeks about the Democratic Socialists of America, as DSA candidates across the country have been running (and winning) in primary races in deep blue districts and cities. So today we’re taking a look at the official DSA platform to find out how the American people feel about some marquee DSA policies. For each policy we’ll examine the DSA platform language and take a look at polling about those issues. Of course, we don’t have time to look at every policy on the website, so we’ve chosen a handful that seem to be getting the most attention.

Healthcare

The official DSA platform language on healthcare is as follows.

Guarantee universal healthcare at no cost to individuals, including complete access to reproductive and gender-affirming care. Expand healthcare access and quality by investing in new public hospitals and clinics. Make all medical education and training public and free.

The first bit here on guaranteeing universal healthcare is commonly referred to as “Medicare for all.” Generally speaking, the idea of the government providing healthcare to every citizen is above water with the American people. For example, in a recent survey by Tavern Research for Zeteo and the Searchlight Institute (Jun. 24 - 26, 2026), 52 percent of likely voters said they would support “the U.S. automatically enrolling every American in a comprehensive government-administered health insurance plan,” while just 19 percent opposed the idea.

However, when you give people more information, support often drops. For example, when the pollsters added information to the question indicating that such a program would be “funded through significantly increased federal taxes,” support fell to 40 percent, and opposition rose to 29 percent. And asked whether they would switch from their current private health insurance plans to a federal program with similar monthly payments, only 30 percent of respondents said they would switch while 70 percent preferred to remain with their current provider.

As for access to reproductive healthcare and gender affirming care provided by the government, it doesn’t seem that the DSA has the public on their side. In a 2024 study by the University of Maryland Program for Public Consultation (Jun. 18 - Jul. 3, 2024), voters were asked to rank how acceptable certain policies are on a scale of 0 (totally unacceptable) to 10 (totally acceptable). Limiting federal funding for abortion was rated as unacceptable (0-4) by less than half — 45 percent — of respondents nationally. Twenty percent rated it as a 5 out of 10 — which the pollster referred to as “tolerable” — and 35 percent rated the policy as acceptable (6-10). Regarding gender affirming care, a 2025 poll from the Pew Research Center (Feb. 10 - 17, 2025) showed only 22 percent of Americans support requiring health insurance to cover gender transition-related medical care, so it seems unlikely that a majority would support the government funding it.

Green New Deal

The official DSA platform language on the Green New Deal is as follows.

Workers must lead our democratic transition to a green future. Pass a federal jobs guarantee. Implement massive investment in sustainable, publicly owned energy and transit infrastructure. Phase out the use of fossil fuels. Reclaim and repair our earth’s resources for all.

Voters have indicated support for a federal jobs guarantee for years. In 2024, Jacobin found that, on average, support for such a program was almost 60 percent. More recently, polling from The Argument/Verasight (May 29 - Jun. 3, 2026) found that 71 percent of voters supported a program “where the federal government would guarantee a job to any adult citizen,” with pollster Lakshya Jain writing that this was “among the most popular proposals I’ve ever tested in any poll.”

Americans also support increased funding for transportation infrastructure, according to a June poll by Morning Consult/The Associated General Contractors of America (Jun. 1 - 7, 2026). In that poll, 47 percent said that they thought federal highway, bridge, and transit funding should be increased, 32 percent said it should be kept the same, and only 7 percent said it should be reduced. When it came specifically to transit, 70 percent said they supported more funding for public transit systems, with just 18 percent opposed.

However, on phasing out fossil fuels, the DSA seems to be going in the opposite direction as the American people. According to a March survey from the Pew Research Center (Mar. 16 - 22, 2026), the percentage of Americans who say that expanding fossil fuels should be a higher priority than expanding wind and solar energy has grown from 20 percent to 42 percent in the last 6 years, while those saying renewables should be a higher priority has shrunk from 79 percent to 57 percent. And slightly more Americans — 38 percent — say the federal government should encourage oil and gas drilling than say it should discourage it — 28 percent.

On the other hand, the Pew survey also finds that Americans strongly support the government encouraging production of wind and solar power, with 65 percent saying so, while just 9 percent say the government should discourage production of renewable energy. So while the DSA may not have the public in hand on phasing out fossil fuels, an increased focus on developing renewable energy could be more popular.

Israel/Palestine

The official DSA language related to Israel and Palestine is as follows.

Recognize the rights of the Palestinian people, including the right of return, the right to resist military occupation, and the right to self determination in a free Palestine with Jerusalem as its capital. End all military and economic aid to Israel. Prosecute U.S. and Israeli leaders responsible for the genocide in Gaza.

American opinion on the Israeli-Palestinian conflict has shifted dramatically, according to decades of polling from Gallup. In the most recent survey (Feb. 2 - 16, 2026) Americans say their sympathies lie more with the Palestinians than the Israelis — the first time this has been the case in Gallup polling going back to 2001. Both independents and Democrats are more sympathetic to Palestinians than Israelis, while Republicans are significantly more sympathetic to Israelis than Palestinians. Favorable views of Israel have fallen to their lowest level since 1989.

Throughout the poll series, Americans have supported an independent Palestinian state. Support has reached a recent high of 57 percent with just 28 percent opposed, higher than at any time since 2003. However, the survey does not specify that such a state would have Jerusalem as its capital, as called for in the DSA platform. While I was unable to find polling with that specific provision, it’s likely that — as we see with Medicare for all — support would decline when specifics were detailed in poll questions.

As to ending military and economic aid to Israel, the DSA seems to have the public closer to their side. In a July poll from Emerson College (Jul. 19 - 20, 2026), 61 percent of likely voters said that the federal government spends too much on aid to Israel. However, completely ending aid may be less popular than simply reducing it. A July survey from Ipsos/The Washington Post (Jul. 8 - 13, 2026) showed that only 29 percent of Americans supported fully ending military support to Israel. Another 26 percent said it should be reduced. So while Americans agree with the DSA that our federal budget spends too much on Israel, it’s not clear that they’d be on board with a complete termination of funding.

Criminal Justice Reform

The official DSA language related to criminal justice and policing is as follows.

Demilitarize police departments, disempower police unions, and redirect funding to public services as steps towards fully abolishing the police and prison system which protects the rich and jails the poor. Treat drug addiction and mental illness as health issues, not crimes. End qualified immunity and empower civilian oversight bodies with real authority over the police.

When it comes to abolishing the police and prison system, the DSA is decidedly on the wrong side of public opinion. A July poll by Change Research (Jul. 13 - 17, 2026) asked about some DSA-related policies, including this. Asked if they supported “ending prisons and policing as they currently exist, including cutting police budgets toward zero and releasing people from prison,” just 13 percent of voters said they supported the policy, while 80 percent were opposed. In a simpler question from an RMG Research/Napolitian News Service poll (Jul. 13 - 14, 2026), 68 percent of voters said they would be less likely to vote for a candidate who supported “defunding the police,” while 15 percent said they would be more likely.

Nevertheless, there are some elements of the DSA platform that are popular with voters. In a 2025 poll from YouGov (Oct. 28 - Nov. 3, 2025), voters favored ending qualified immunity by 23 percentage points, 54 to 31. And policies related to police oversight had enormous support, including making it easier for the federal government to charge officers for using excessive force (supported by 69 percent) and creating a federal registry for complaints against officers (supported by 72 percent).

The YouGov survey also showed mixed opinions on redirecting funding to public services. Asked whether they would support “budgeting less money for your local police department and more for social services (for instance, funding social workers and mental health professionals),” 46 percent of respondents said they were in favor, while 38 percent were opposed and 16 percent were unsure. But as with the more recent surveys, opinion was clear on defunding the police: the policy was underwater by 40 percentage points.

There are, of course, many other policies in the DSA platform that have varying levels of support among the public. For example, their call to invest in new, publicly-owned housing is generally popular, while their call for reparations to Black Americans is not. Many of their policies related to electoral reform would find favor with the American people, such as ending Citizens United and getting money out of politics. Others might not, like ranked choice voting. Still others might be more mixed, such as making Washington, D.C. a state.

Opinions on the DSA platform reflect the complexity of how Americans view policy. The details underlying each proposal span the gamut of popularity, with some aspects overwhelmingly favored among the electorate and others deeply underwater. How these candidates ultimately fare in future elections may hinge on which policies they emphasize, and whether they are able to jettison the least popular ideas in order to win over voters.

Share

Other polling nuggets

• A poll from Politico/Public First (Jul. 13 - 15, 2026) found that voters across the political spectrum overestimate how much gas prices have increased since Trump took office. While gas prices have, on average, increased just $0.67 since Trump’s inauguration, according to Politico, 23 percent of respondents say gas has risen in their area by $3 or more. Another 17 percent said prices have risen by between 2 and 3 dollars, and 25 percent said they’ve risen by between 1 and 2 dollars. Of course, respondents may be having a difficult time with the timeline here. Gas prices fell in the first year of Trump’s administration before rocketing up due to the conflict in Iran, so respondents may be comparing current prices to the more recent lows rather than the beginning of Trump’s term.

• Opposition to AI data centers continues to grow, according to the latest Emerson College poll (Jul. 19 - 20, 2026). A significant majority, 63 percent, of likely voters said they would oppose data centers “being built in or near your community,” up from 42 percent in December 2025. Just 27 percent said they would support the centers, down from 33 percent in December. Opposition is strongest among Democrats, with 77 percent opposed, compared to 66 percent of independents and 47 percent of Republicans.

• Nearly one in three Americans are sharing their bed with at least one furry friend according to a YouGov poll (Jul. 24, 2026). Thirty-one percent said they sleep with at least one pet, including 10 percent who said they sleep with 2 or more. And unlike the divisive issues of our time, sleeping with a pet is bipartisan: 32 percent of Democrats, 30 percent of independents, and 30 percent of Republicans are all sharing their beds with at least one furbaby. Sweet dreams, folks.

Share

Polling averages update

All numbers are as of 1:00 PM Eastern on July 24, 2026.

2026 U.S. House generic ballot

The Democratic margin in the 2026 U.S. House generic ballot rounded to 6 points this week, up from 5 points last week. Currently, 49.3 percent of likely voters intend to vote for a Democrat in November, compared to 43.7 percent who plan to vote for a Republican.

Trump’s job approval

President Trump’s approval rating among adults decreased this week, from 37.1 percent approval to 36.8 percent. At the same time, disapproval rose from 58.9 to 59.4 percent.

Paid subscribers to 50+1 get access to premium analysis, plus sortable tables and complete data access on our polling website. If you want to follow the 2026 cycle with the best data at your fingertips, become a paid subscriber.

Upgrade to a paid subscription today

In the meantime, here’s a fresh look at what the polls are saying about the battle for Congress today, according to our seat-level “nowcast” model. You can read more about how that model works here.

Paid subscribers to 50+1 get access to premium analysis, plus sortable tables and complete data access on our polling website. If you want to follow the 2026 cycle with the best data at your fingertips, become a paid subscriber.

Upgrade to a paid subscription today

And if you’re in the market for either the raw polling data used in this analysis, pre-calculated polling averages, or the data behind our upcoming election forecast for each House and Senate seat, check out our API subscription and get in touch about enterprise-level data access.

Explore custom data access

A polling averages update

Let’s start with the national picture. According to our published average, Democrats currently lead the generic ballot 49.2% to 44.0% — a 5.2-point edge, down from roughly six points at our last nowcast update. The movement is almost entirely on the Republican side of the ballot, with undecided voters gravitating into the GOP voter pool while the Democrats’ vote share has sat roughly flat since mid-May.

When we consider polling alone, the race-level picture still tilts blue. Democrats currently lead in four Republican-held Senate seats: North Carolina (D+5.5), Maine (D+1.9), Ohio (D+0.6), and Texas (D+0.6). Make that five if you include Nebraska, where independent candidate Dan Osborn — who says he would caucus with neither party, but whose seat we count toward Democratic control — leads by five. (Be careful with this one; only one survey has been released in this race since February, so there are wide error bars around Osborn’s lead.)

Additionally, Democrats are polling within a point in two more Senate races: Iowa (R+0.9) and Alaska (R+0.9).

On the Democratic side of the map, New Hampshire (D+7.3), Georgia (D+7.1), Minnesota (D+5.8), and Michigan (D+0.6) round out the competitive seats; Florida (R+7.9) remains the GOP’s most comfortable lead in races for which we have polls.

Maine deserves a quick note. It now looks almost certain that Troy Jackson, former President of the Maine Senate, will be the new Democratic nominee to take on Susan Collins in November. We have gathered four polls testing support for his campaign: Wedgewood Polls (July 4–6) had him up 48–43, PPP (July 7, conducted for Platner’s campaign before he dropped out) up 49–44, Z to A Research (July 7–8, sponsored by 314 Action, who was supporting Nirav Shah to replace Platner) down 47–49, and the first high-quality nonpartisan survey from the University of New Hampshire (July 15–20) has Jackson up 49–46. Our adjusted average puts the race at Jackson +1.9 — tighter than the raw toplines because of house-effect and sponsor adjustments, but a good place for a challenger to sit against the incumbent Collins, who has made a career of beating expectations.

In the House, ten districts have enough polling for an average. Arizona’s 6th (D+0.1) and Michigan’s 4th (R+0.1) are dead heats; Pennsylvania’s 8th (R+0.7), Wisconsin’s 3rd (D+1.4), Maine’s 2nd (R+2.3), and Michigan’s 7th (R+2.6) are all inside the margin of error.

Share

State of the nowcast

If the election were held today, our nowcast model gives Democrats a median of 50 seats in the Senate (counting Osborn’s) and a 46% chance of controlling the chamber, down a few points from two weeks ago. The Democratic hold on the House has also slipped slightly, now at a median of 227 Democratic seats and a 79% chance of the majority (82% at our last update) with district polls generating an implied national House vote of about D+6.7 (rosier than the current generic ballot average).

According to the joint simulations from our model (which, unlike others, simulates House and Senate elections simultaneously), Democrats sweep both chambers about 44% of the time, Republicans hold both about 19%, and in roughly 38% of simulations the parties split control.

The big movers in the Senate are: North Carolina, where Democratic candidate Roy Cooper has slipped from D+9 to D+5.5 and New Hampshire, which has sunk from D+9.7 to D+7.3. And our model now has more uncertainty about what will happen in Maine, given the lack of polling in the new matchup between Jackson and Collins.

If the election were held today, we are fairly confident the Democrats would pick up the House and have an outside chance at taking the Senate, too. Yet there is a long way to November. A full forecasting model for the House and Senate is coming out soon here at 50+1. It will be accessible here on our Substack and website.

FiftyPlusOne publishes daily updating polling averages for every competitive House and Senate race, plus the raw data behind them, at FiftyPlusOne.news. Enterprise subscribers can get all of that data and more via our data API.

Explore custom data access

Upgrade to support 50+1

On Thursday, July 16, 2026, President Donald Trump gave a primetime address in which he discussed American election integrity, a topic he has focused on since his election loss in 2020. In the speech, Trump said that U.S. elections are vulnerable to foreign interference, particularly from China, and repeated unproven allegations of widespread voter fraud, particularly in the 2020 election. One of the president’s claims is that the Chinese Communist Party “hacked” U.S. elections by collecting person-level records of registered voters called “voter files” which are free to download in many states (for example, here is North Carolina’s).

To be clear, despite years of trying, the Department of Justice has not found evidence of widespread systemic voter fraud in the 2020 election. Trump and his supporters have lost dozens of lawsuits alleging misconduct in the election, and independent analyses have repeatedly shown no evidence of systemic fraud.

But Trump’s repeated insistence that American elections are vulnerable may be impacting how voters think about our democracy, and whether or not it is secure.

Since 2017, researchers at Bright Line Watch have been asking both the public and political science experts to rate U.S. democracy on a scale from 0 to 100. Until the 2020 election, ratings of democracy were fairly stable among the public. Republicans typically rated our democracy around 60, and Democrats around 50. But in the wake of the 2020 election, those ratings flipped: Republicans’ evaluation of democracy fell below 50, and Democrats’ evaluations rose to above 60 for the first time in the survey’s history.

Notably, though, expert ratings of democracy didn’t budge in that time period (though they were higher before 2017 in other datasets). From 2017 through November 2024, experts consistently rated U.S. democracy between 60 and 70 on the 100-point scale. But through Trump’s second term, that rating has plummeted. Experts now rate U.S. democracy at close to 55 on the 100-point scale, and ratings among the public have also dropped from the mid-50s to the high 40s/low 50s. This has been driven by a collapse among Democrats, whose ratings of democracy plummeted from the mid 60s prior to the 2024 election to the mid 40s today.

Other surveys also reveal a lack of confidence in U.S. democracy. In a recent YouGov poll, majorities or near-majorities of Americans said they believed the 2026 elections would be tampered with in a variety of ways, including tampering with vote counts, voter fraud, and failure to properly count ballots.

Despite believing that various malfeasance will occur, however, Americans don’t think it will change the results of the elections. And when you drill down, there’s broad differences between the sorts of issues Democrats and Republicans are concerned about. For example, 71 percent of Republicans think there will probably be voter fraud in the 2026 elections, while just 34 percent of Democrats do. On the other hand, 66 percent of Democrats say that some people who are eligible to vote will be prevented from doing so; just 33 percent of Republicans agree. The one issue on which everyone seems to agree? That false information will be spread about the elections: 76 percent of Americans overall say that will probably occur, including 79 percent of Democrats and 76 percent of Republicans.

Trust in election administration is also low. In a June Marquette Law School poll, just 38 percent of Americans said they had a great deal or quite a lot of confidence in “the people who administer elections in your state;” another 34 percent said they had some confidence, and 28 percent said they had very little or no confidence at all in local elections officials. This lack of trust is especially partisan: 29 percent of Republicans and 22 percent of independents said they had a great deal or quite a lot of confidence in local elections officials, but 53 percent of Democrats had confidence in election administrators.

On the other hand, in the same poll, Democrats were far more pessimistic about our democracy than Republicans. While 52 percent of Americans overall said they were very or somewhat optimistic about our country’s future as a democracy, just 41 percent of Democrats said so, compared with 66 percent of Republicans.

The lack of confidence in local elections administrators is particularly concerning, especially as Trump has attempted to federalize some elections administration work. According to the latest Bright Line Watch survey, 93 percent of political science experts said that calls to nationalize voting were a threat to democracy, including 85 percent who rated it as a serious or extraordinary threat. Eighty-eight percent said that the DOJ suing states for access to their voter lists was a threat to democracy.

However, it doesn’t appear that Trump is the only factor influencing how Americans feel about election integrity. Before the 2020 election, Bright Line Watch found that across the political spectrum, confidence that votes would be counted as voters intended was less than 70 percent. And in the years since then, confidence in votes being counted as voters intend has been highly polarized, and often tied to whether or not voters’ own party did well in the elections. Confidence in votes being counted accurately has been falling among Democrats since the 2024 election, for example, while Republican confidence spiked in December 2024.

Trump’s continued insistence that American elections are insecure has made election integrity a partisan issue, where it may not have been one in the past. However, while the issue makes political science experts and close poll watchers nervous, it may not be as big of a deal to the American public. In a June 2026 survey from Morning Consult and the Cato Institute, voters were asked which issues are the “greatest threat to the United States remaining a free republic.” Given a list of options and asked to choose up to three, “people refusing to accept election results” was the second least often chosen, with 10 percent selecting it (the least concerning was “falling birth rates” at six percent). The options chosen most frequently as threats to our free republic were corruption at 30 percent and “politicians ignoring the Constitution” at 26 percent. And perhaps that should be more concerning for Trump: in February, over half of Americans said the president is corrupt.

Share

Other polling nuggets

• A majority (55 percent) of Americans think the U.S. should reduce or end military support to Israel, according to a Washington Post/Ipsos poll (Jul. 8 - 13, 2026). Twenty-nine percent said the U.S. should end its military support of Israel, and 26 percent said it should reduce support. Just eight percent of respondents said the U.S. should increase military support for Israel, and a plurality, 32 percent, said that support should be kept at current levels. Democrats are especially likely to say support should be reduced or ended: 72 percent of Democrats said so, while 59 percent of Republicans said it should be increased or kept at current levels. Republicans that identify as MAGA were the most likely to support continued funding of Israel’s military, with 69 percent saying support should be increased or kept steady.

• Patriotism looks different around the world, according to a recent Politico/Public First survey (Jun. 14 - 17, 2026). While all six countries polled (the U.S., U.K., France, Canada, Spain, and Germany) rated voting in elections as the most patriotic act, things diverge from there. In the U.S., voters rated serving in the military and displaying a national flag as the next two most patriotic acts. In France, Spain, and Canada, buying products made in respondents’ home countries were rated highly, and both Spain and France rated paying taxes as an act of patriotism. In the U.K. and Germany, singing the national anthem was rated highly as an act of patriotism, and respondents in the U.K. said that supporting their country in international events like the World Cup was an act of patriotism.

• After Senator Lindsey Graham’s death and Senator Mitch McConnell’s recent hospitalization, voters support more transparency about politicians’ health. In a Tavern Research survey (Jul. 13, 2026), given some information about these circumstances, 62 percent of voters said they would support new Senate rules requiring Senators to update the public about their health status, while just 20 percent were opposed. Support was bipartisan, with 51 percent of Trump voters and 78 percent of Harris voters in support of new rules requiring health disclosures.

• The World Cup final matchup between Argentina and Spain is set for this Sunday, July 19, 2026. In a YouGov poll (Jul. 16, 2026), 36 percent of Americans said they plan to watch the match, while 56 percent said they won’t. And America doesn’t seem to prefer one team or the other: 20 percent said they wanted Argentina to win, while 23 percent said they wanted Spain to win. Forty-four percent said they didn’t know who they wanted to win, and 14 percent said they wanted neither team to win (this group will, ultimately, be disappointed).

Subscribe now

Polling averages update

All numbers are as of 12:00 PM Eastern on July 17, 2026.

2026 U.S. House generic ballot

The Democratic margin in the 2026 U.S. House generic ballot rounded up to 5 points this week, down from 6 points last week. The percent of likely voters indicating they will vote for a Democrat stayed mostly the same, going from 49.3 to 49.2 percent this week, but the percent saying they will vote for a Republican slightly increased from 43.5 percent last week to 43.8 percent this week.

Trump’s job approval

President Trump’s approval rating among U.S. adults remains underwater by 22 points this week, with very little movement compared to last week — the percent of adults who approve of the president remains the same at 37.1 percent, while disapproval has fallen only a tenth of a percentage point.

Paid subscribers to 50+1 get access to premium analysis, plus sortable tables and complete data access on our polling website. If you want to follow the 2026 cycle with the best data at your fingertips, become a paid subscriber.

Subscribe now

It’s summertime, which means the Supreme Court is in the news again.

The court’s overall approval rating has been abysmal for years now, and according to the latest YouGov/Economist poll conducted earlier this week (Jul. 3 - 6, 2026), most of the justices’ personal favorability ratings are in the gutter as well. Only the three liberal justices have positive net favorability ratings, while the six conservative members are underwater to varying degrees.

Part of this gap is partisan: In every case, Democratic respondents in the poll were more likely to express an opinion about the justices than Republican respondents. But among independents, the least likely group to express an opinion, we see the same pattern: Every Democratic-appointed justice has a net-positive favorability rating, and every Republican-appointed justice has a net-negative rating.

The decisions they authored this year probably won’t do much to help their numbers. The court closed its term last week with highly consequential rulings on birthright citizenship, transgender athletes, and campaign finance law, and it might come as little surprise that many of the most controversial decisions this year were unpopular with the public.

According to SCOTUSPoll, an annual survey conducted by researchers at the University of Texas and Harvard University, a majority of the public agreed with the court’s decision in only six of 14 major cases decided this term. On average across the cases that SCOTUSPoll tracked, less than half (48 percent) of the public agreed with the Supreme Court’s eventual ruling.

It’s certainly not all bad news for the court; some of the biggest decisions of the term, including those surrounding birthright citizenship and President Trump’s authority to impose tariffs on other countries, were also the ones most supported by the public.1 But many others, including decisions surrounding campaign finance law and a pair of 2nd Amendment challenges to gun regulations, were deeply unpopular.

Opinion on a lot of these cases isn’t particularly strong, with over half falling between 40 and 60 percent support, so we should probably approach them with a healthy amount of caution. Polling about Supreme Court cases can be especially susceptible to question wording bias, since most Americans are often unfamiliar with the details of each case and don’t have strong opinions yet. Because of that lack of familiarity, public opinion is often quite malleable, particularly during the buzzy period just after a ruling is handed down.

The SCOTUSPoll survey was conducted in the spring, before most of the major cases of the term were decided. That means that public sentiment could have shifted, perhaps dramatically, in the months since, as voters were exposed to a barrage of news coverage and reactions from politicians in the wake of each decision. When support or opposition for a ruling is hovering in the 40s or 50s, there’s a good chance that either the question wording or the news coverage surrounding a decision could significantly impact public opinion. It’s probably a good idea not to draw too many sweeping conclusions from the results for those “squishy” cases where voter sentiment isn’t as clear-cut.

But beyond the specific cases of this term, the public also seems uneasy with the amount of deference the Supreme Court has shown to Trump and conservative causes. In the YouGov/Economist poll, 45 percent of respondents said the court’s recent decisions had given the president too much power, versus 9 percent who said they’d given him too little power, and 29 percent who felt they had given him “about the right amount of power.”

In another YouGov poll from May, 61 percent of adults said that the justices “often let their own personal or political views influence their decisions,” and in a separate question just 25 percent said that the court was impartial in its decisions, compared to 49 percent who said they tended to favor one group or another.

The Court’s term is done, so we probably won’t be hearing much about them for at least a few months. But there are rumors swirling about potential retirements, so there may be more to come soon. I certainly wouldn’t want these numbers hanging over my head as I start my summer vacation.

Share

Other polling nuggets

• On Wednesday, Maine Democratic Senate candidate Graham Platner dropped out after new allegations of sexual assault surfaced. The focus now shifts to deciding who will replace Platner on the ballot, which is ultimately up to the state Democratic party. A number of candidates have thrown their hats into the ring, including former state Senate President Troy Jackson, Secretary of State Shenna Bellows, and epidemiologist Nirav Shah, all of whom ran unsuccessfully last month in the Democratic primary for governor.
  
  The limited public polling we have so far shows Jackson may be best positioned to defeat Republican Sen. Susan Collins in November, but it’s a bit too early to know anything for sure. The day before Platner dropped out, PPP conducted a flash poll for his campaign, testing various hypothetical head-to-head matchups against Collins to see if the oyster farmer was still viable — and if not, who might be. The results showed Jackson performing the best of the group, and he was the only candidate tested who polled ahead of Collins (by five points, 49 percent to 44 percent). Bellows and Shah were tied with Collins, while Platner, Gov. Janet Mills, and former U.S. Senate candidate Jordan Wood were all trailing the GOP senator. A Wedgewood Polls survey conducted right before news of the latest allegations became public also showed Jackson ahead by 5 points, 48-43 percent.
  
  But a Z to A Research poll released yesterday and conducted for 314 Action, a super PAC supporting Shah, showed a more mixed bag. Of the three candidates tested, Jackson actually performed the worst, trailing Collins by two points, while Bellows was tied and Shah was up a point. Ultimately, though, all three of the matchups were well within the margin of error and should be considered a statistical tie.

• On the heels of a controversial red card given to U.S. striker Folarin Balogun during a World Cup match against Bosnia-Herzegovina (and its subsequent last-minute reversal), the U.S.’s round-of-16 match against Belgium became the most-watched soccer game in American history. Earlier this week, HarrisX/2Way asked Americans what they thought about the red card debacle, and among the half of respondents who had an opinion, the results were resounding. Thirty percent of Americans said Balogun did not deserve the red card, while just 16 percent said he did (54 percent said they weren’t sure). Thirty-two percent said FIFA was right to reverse the decision, while 19 percent said the red card should have stood. But following reports that Trump personally lobbied FIFA President Gianni Infantino to suspend the red card, most Americans seemed to think he crossed a line. Forty-seven percent said that Trump’s actions weren’t appropriate, while 24 percent said they were and 28 percent weren’t sure. All the drama surrounding the red card and reversal didn’t end up mattering much in the end, though: Belgium handily dismantled the U.S. team 4-1.

• Support for Israel continues to slide among Democrats and the public as a whole, according to an AP-NORC poll conducted last month. According to the poll, the share of respondents who said the U.S. “wasn’t supportive enough” of Israel fell from 25 percent in 2024 to 18 percent in June. At the same time, the share of adults who said the U.S. was “too supportive” of Israel in its conflict with Palestinians inched up to 40 percent from 37 percent in January 2024. Among Democrats, though, the increase was more pronounced, going from 45 percent in 2024 to 58 percent in June.

Polling averages update

All numbers are as of 11:00 AM Eastern on July 10, 2026.

2026 U.S. House generic ballot

The Democratic margin in the 2026 U.S. House generic ballot remained stable this week, rounding to 6 points. Currently, about 49.3 percent of likely voters say they plan to vote for a Democrat in November, while 43.5 percent say they plan to vote for a Republican.

Trump’s job approval

President Trump’s approval rating among U.S. adults barely budged this week, with his approval staying the same and disapproval dropping 0.4 percent. An average of 37.1 percent of Americans approve of the president, while an average of 59 percent disapprove.

Paid subscribers to 50+1 get access to premium analysis, plus sortable tables and complete data access on our polling website. If you want to follow the 2026 cycle with the best data at your fingertips, become a paid subscriber.

Upgrade to a paid subscription today

Editor’s note: We did the number-crunching for this article before Politico published a new account of sexual assault allegations against Maine Democratic Senate candidate Graham Platner. It now looks like Platner will imminently end his campaign, though a decision has not been announced.

Because he has not yet dropped out, however, we have decided to go ahead with this article that includes polls of a Platner vs Susan Collins general election. The topline numbers you see here will stand if Democrats nominate a candidate who polls at least as well as Platner did against Collins prior to the allegations.

It’s a midterm election and the polling average to watch this early is the generic ballot average. That’s the question that asks voters who they would cast their ballot in their local U.S. House district if the election were held today.

Democrats lead our average on this question by roughly six points. That’s a slight decrease from their position a month ago, and just below where the party was at this point in 2018 — when the party won 235 seats in the House in what pundits called a “blue wave” election.

But the generic ballot is a national number. While it is predictive of the national popular vote for the House, actual seat-level outcomes can diverge sharply from what is implied by the national environment (incumbency, experience, and candidate effects can matter a lot — ask Richard Mourdock). Nobody casts a vote in a national election — control of Congress gets decided seat by seat, in 435 districts and 33-odd Senate races.

This means the generic ballot can only approximate what voters are saying in individual races. Which invites an obvious question: what do the actual seat-level surveys say?

Lucky for readers, 50+1 has all the polls of these individual races and can do the math! In this piece I preview our upcoming general election polling averages for the House and Senate and compare the implied national environment from these surveys against the generic ballot. I also take stock of the current race for the Senate using the portion of our upcoming election model that estimates opinion today in each competitive Senate race — something we’re calling a “nowcast” of the election. More details below.

Paid subscribers to 50+1 get access to premium analysis, plus sortable tables and complete data access on our polling website. If you want to follow the 2026 cycle with the best data at your fingertips, become a paid subscriber.

Upgrade to a paid subscription today

And if you’re in the market for either the raw polling data used in this analysis, pre-calculated polling averages, or the data behind our upcoming election forecast for each House and Senate seat, check out our API subscription and get in touch about enterprise-level data access.

Explore custom data access

Two ways to measure the political environment

To be clear, we like the generic ballot. It’s one of the best early indicators we have for the House, and it gets more predictive as Election Day nears. Our average of D+6 today should be taken seriously as a robust historical predictor of the eventual national environment according to polls on Election Day.

Historically the out party tends to gain a few points in the generic ballot from now to November, but other than that if you were predicting House elections and had to pick between nothing and the generic ballot, you would take the generic ballot every time.

But seat-level surveys are a much better predictor of seat-level results. They ask about actual votes in a given race — not which party voters they would prefer in the abstract, but which candidate they are actually going to choose in a real election.

Right now we’ve got polling averages running behind the scenes here at 50+1 in 25 of this cycle’s contested races — 9 House and 16 Senate contests. Let me briefly explain how our polling averages work. (If you’re a real wonk, the full methodology lives here.)

Share

State polling averages

Our seat-level averages are built the same way for every race, House or Senate, in four steps.

1. Decide which polls count. First we filter all the polls for a given race to exclude those of hypothetical matchups: surveys testing, for example, Mike Rogers vs. Gretchen Whitmer in Michigan are discarded (the latter is not running for Senate). Then, we limit each poll to one sample — if a pollster tests the race among all adults, registered voters, and likely voters, we prefer the latter (and we prefer RV to adults). Then, to keep our average recent, we filter to the last 90 days of polling (but if polling in that window is too thin we reach further back until we have the five most recent polls in each race).

   
   Some other minor restrictions apply and are noted here: for example, we do not include “polls” that use MRP or other statistical models to turn national samples into state estimates. (We wrote about this previously; we think such techniques are cool and valid from a prediction standpoint, but they break the logic and statistics of aggregation.

   
   As a final note, we don’t publish an average at all unless a race has at least three polls from at least two different pollsters — one campaign’s internals do not get to be an “average.”

2. Adjust each poll for when it was taken. Taking the available pool of surveys in a given race, we then apply what we call a “trendline adjustment” to each result to account for any movement in the national political environment since a poll was fielded. We do this using the day-to-day movement in the generic ballot: if Democrats gained, say, two points nationally since a poll was conducted, we increase the Democratic candidate’s share of the vote in the survey by two points. This brings every survey up to the present — borrowing the national trend, while the level of each race comes from its own polls.

3. Adjust for who did the polling. Different polling firms can produce persistently, systematically different numbers from other firms because of the choices they make in their data-generating process. Some pollsters lean a couple of points toward one party across everything they field — up to 6 or 7 points on margin in some cases. We accordingly estimate so-called “house effects” for each pollster by comparing its surveys in each race to the average in that race. We do this for each pollster across all the races in a cycle. We adjust polls from each pollster by subtracting their house effects, which are estimated independently for Democrats and Republicans (some pollsters simply have an elevated Don’t Know share).

   
   We estimate a similar “mode effect” for all polls that share the same methodology, using a similar comparative process.

   
   Finally, polls sponsored by a campaign or other partisan organization are additionally shifted toward the sponsor’s opponent by one point, and get less weight in the average.

4. We put everything on a likely-voter basis. Comparing different’ pollsters estimates of Democratic/Republican support among different populations in the same race, we estimate the residual partisan difference between likely voters, registered voters, and all adults. Individual polls are then shifted toward the likely voter universe by subtracting these difference between pollsters.

5. We decrease the weight on polls from firms that release data frequently, and from those that are of lower quality, and increase weight by sample size. We apply the former by down-weighting each poll from a pollster in a given race by the square root of the number of polls it conducted in the last three weeks, and the latter by down-weighting according to pollster reliability scores estimated by FiveThirtyEight ahead of the 2024 election. We aim to calculate our own, more up-to-date pollster reliability scores soon. We apply a weight to polls based on their sample size.

There are a couple other bells and whistles described on our methodology page. For example all of these effects enter a Bayesian model that accounts for uncertainty in each parameter, including the estimates of a firm’s house effect, its downweight amount, etc. And we also account for additional non-sampling error that can push a poll off course, which tends to make our averages more robust to outlier surveys than the averages you might see at other outlets.

At this point, what comes out of our series of models is a trendline-adjusted, house-effect-corrected, likely-voter average for each House and Senate race with enough polls. We also put more weight on recent polls, higher-quality samples, and pollsters who don’t “flood the zone.”

Here’s what the current state of key Senate races looks like, according to a pre-release version of our polling averages:

Note a tight cluster of competitive seats in the R+5 to D+5 zone. These are the core competitive seats that will decide control of the Senate majority this November. Democrats need to win in four states currently under Republican control, plus hold on to their current seats in Michigan and Georgia, to take the chamber.

More on races, and the state of play, in a moment.

Subscribe now

The national environment, according to House and Senate polls

Armed with our polling averages in every Senate and House seat, we can now calculate the implied national environment from subnational polls. Do contest-level surveys agree with the generic ballot? Are Democrats heading for a big win in the House, and a potential flip of the Senate?

The method for turning those averages into a national reading is simple. Every seat has a partisan baseline — how it tends to vote relative to the country. We estimate that from the last two presidential elections (weighted 85% to 2024 and 15% 2020) and center it so a perfectly tied nation equals zero. We call this a seat’s “partisan lean.”

We can thus take a race’s polling average, subtract its lean, and whatever’s left is the implied national environment given that state’s results.

Here’s a visual comparison of current polling averages to each state/seat’s benchmark partisan lean. Note Democrats are generally polling ahead of their benchmark in each seat — according to our estimates — by an average of 7 points.

Some contests really stand out here. For one, the Senate race in Nebraska where independent candidate Dan Osborn is hoping to knock off independent Republican Pete Rickets, is much more competitive than you’d expect based on the state’s ruby-red partisan lean. Maine goes the other way: it votes about 9 points to the left of the country as a whole, but the Democratic challenger there, Graham Platner, is within the margin of error with Republican incumbent Senator Susan Collins.

Then there’s a cluster of seats in the neutral zone that look to have shifted about 10-12 points left: Ohio, Iowa, Alaska, and Texas. The Lone Star State is a good example. The polls there have the Democratic candidate, James Talarico, within about a point of the Republican State Attorney General Ken Paxton — a state that normally votes about 12 points redder than the country.

For a Democrat to be running even in Texas, the national mood has to be somewhere around D+13. Yet one race can be noisy — candidate quality, incumbency, sponsor effects, and local conditions all matter. But repeat the same calculation across 25 races and the idiosyncrasies start to wash out.

If you control for these state- and candidate-level idiosyncrasies (using something called a multilevel or random effects model), the implied national environment using Senate polls today is D+8.1, and using House polls it’s D+6.2. The combined implied national environment is D+7.4.

Subscribe now

What a D+7-to-D+8 environment means for the House and Senate — introducing the 50+1 polling “nowcast”

All of these polling averages are building blocks for something bigger: our 2026 election forecast, which will be out in the coming weeks. But before a model can say anything about what might happen in November, it has to solve a more basic problem — how does the polling we have today translate into seats on the ground today? Polls exist for a couple dozen races, but control of Congress will be decided across all 435 House districts and every Senate seat on the ballot. Something has to fill in the map.

That “something” is our “nowcast” model. It’s a statistical model (a Bayesian measurement model, for the wonks) that treats each race’s true state of opinion as an unknown quantity and every poll as a noisy reading of it. The model anchors the national environment to the generic ballot and then — similar to what we’re doing here — lets the race-level polls place each contest relative to that anchor. Where a seat has polling, the polls dominate, weighted by their sample sizes and adjusted as described above. Where a seat has none — which is most of them — the model imputes a result from the seat’s partisan lean, its incumbency status, and the predicted national environment based on the generic ballot and those other seat-level polls.

Then we simulate the outcome of the election at the chamber level. The model draws thousands of plausible values of the popular vote “if the election were held today,” and then messes with the state/district map the way real polling errors work. The model simulates a random national miss that shifts every seat in the same direction (think 2020), plus race-specific errors that hit each contest independently (like in 2022 and 2024 when Democrats beat expectations in competitive seats in the House). There are a couple other sources of error too, including potential polling bias in each contest and, at the House level, error that is shared across seats with similar demographics and political geography. (In 2018, Democrats beat the polls in suburban seats.)

After simulating thousands of election outcomes, we can then count up the seats each party wins in each simulation. This gives us a distribution of seats they would win if the election were held today. From that we can derive a median, an uncertainty interval, and a probability that each party controls each chamber — again, if the election were held today.

Here’s what that chart looks like:

Again, this is a hypothetical exercise — purely a set of statistics we can derive from the machinery of our eventual forecasting model without actually addressing the forecasting part of it. That is, of course, the hard part: how things change between now and November is uncertain. But the nowcast serves two crucial purposes: figuring out what the election looks like in seats where we don’t have polling and measuring the historical reliability of generic ballot, House, and Senate polling on election day.

Here is how the nowcast has changed over time:

We are reasonably sure that if the election were held today, Democrats would take the House pretty comfortably with a median of 226 seats — well past the 218 they need for the majority. They carry the Senate in about half of our simulations.

In the Senate, Democrats would be expected to win a median of 51 seats (a mean of 50.5) if the election were held today — exactly enough for a Senate majority — IF you treat a victory for independent candidate Dan Osborn in Nebraska as an effective Democratic win. But this “if” deserves unpacking, because Osborn says he won’t caucus with either party. We count a win for him as an effective Democratic seat, however, because it deprives Republicans of a vote for their caucus. If Democrats win 50 seats themselves and Osborn takes Nebraska, Democrats can organize the chamber 50–49 without him so long as he merely declines to vote for a Republican majority. And while Vice President JD Vance breaks tied votes in the Senate, Osborn can cast an absent vote in proceedings for adopting rules and committee appointments. If you don’t count Osborn’s win as a Democratic seat, the party’s chance of winning the Senate in the nowcast falls from 50% to 41% — still what we could consider a “toss-up.”

However if you count him, the conclusion is the same: the Senate is a toss-up if the election were held today and all you looked at was polling. A surprising share of the paths to a Democratic (or Republican!) majority currently run through a state Republicans carried by twenty points just two years ago.

Share

From nowcasting to forecasting

There are limits to the nowcast, however. Ultimately we think its useful mostly as a stepping stone to a full-blown forecast.

For one thing, while we trust polls, they are not the only information we have at our disposal for predicting election outcomes. States tend to vote rather predictably over time, so a full-throated election model should take into account not just the polls but the traditional partisan edge in a seat. Further, we know from modeling all congressional elections since 1998 that candidates who have held office before and fundraise well typically beat expectations. Republican Senator Susan Collins, for example, won re-election in Maine in 2020 by 8 points while Joe Biden carried it by nine.

There’s also the uncertainty inherent in predicting an election that occurs in the future, not today, to take into account. For that we study how polls have moved historically over the course of past congressional elections, at both the national and race level. We consider candidate fundraising and other data that historically has been predictive of election outcomes. Finally, we put a lot of time into figuring out how to integrate expert race ratings from the Cook Political Report, Sabato’s Crystal Ball, and Inside Elections, which is not a trivial task.

A full forecasting model for the House and Senate is coming out soon here at 50+1, accessible via our Substack and website. Subscribers to this newsletter can expect updated nowcasts in the meantime.

FiftyPlusOne publishes daily updating polling averages for every competitive House and Senate race, plus the raw data behind them, at FiftyPlusOne.news. Enterprise subscribers can get all of that data and more via our data API.

Explore custom data access

Subscribe now

A recent headline from The Hill caught my eye this week: “Almost 7 in 10 say Supreme Court should keep birthright citizenship in place: Survey.” The survey the article is referencing was released this week by Quinnipiac University, and showed 69 percent of respondents supporting birthright citizenship, while 27 percent were opposed.

I found this startling, because it doesn’t really jibe with the other polling we have on this issue. This turns out to be one of those types of questions where how a pollster writes the survey matters a lot to how people view the issue. So this week on The Trendline, we’re going deep on birthright citizenship, how pollsters ask about it, and what we can glean about Americans’ opinions on the issue.

To set the table, a bit of history. The first sentence of the 14th Amendment to the U.S. Constitution states “All persons born or naturalized in the United States, and subject to the jurisdiction thereof, are citizens of the United States and of the State wherein they reside.” In various court cases, this sentence has been clarified; in 1898, the Supreme Court ruled that the sentence applies to anybody born on U.S. soil, regardless of the immigration status of their parents. (There are some rare exceptions, such as for diplomats and foreign royalty.)

Despite various legal challenges, this 1898 precedent has stood ever since. But in January 2025, President Donald Trump signed an executive order seeking to limit birthright citizenship only to babies born to parents that have either citizenship or permanent legal status in the United States. This order was quickly challenged in court and is one of the final Supreme Court cases that is yet to be decided this term. (We expect a decision next week.)

As with any hot-button issue before the court, we’ve had plenty of polling on whether people agree with the Trump administration here or if they prefer to keep birthright citizenship in place. By and large the polling shows that people support birthright citizenship as currently understood, but the margins vary significantly by pollster. (Note: these are not all the polls we know of that ask about birthright citizenship but a sample to demonstrate the range).

Of course as with many policy issues, a lot of the differences come down to how pollsters word the question. The pollsters that found lower support generally presented the question in a framing that emphasized the legal status of the parents. For example, the YouGov poll, which had the lowest level of support for birthright citizenship at a margin of just 9 percentage points, specified in the question wording that the issue is about undocumented immigrants. They asked if respondents would support “limiting birthright citizenship by no longer giving U.S. citizenship to children born in the U.S. either to undocumented immigrants or to some temporary visitors.” The Data for Progress poll also mentioned the legal status of the parents in the question wording.

The Ipsos/Reuters poll, which also showed relatively low support, framed this question more neutrally: “Do you support or oppose ending birthright citizenship, which makes anyone born in the U.S. a citizen?” But a different type of polling issue was at play here, known as “priming.” Immediately before this question, the survey asked respondents if they “think it’s a good thing or bad thing that children born in the U.S. are U.S. citizens by birth, regardless of the legal status of their parents?” So even though the birthright citizenship question is framed neutrally, respondents were prepared by the pollster to be thinking about the legal status of the parents before the question was presented.

On the other hand, the Quinnipiac survey, which found the highest support for birthright citizenship with a margin of 42 percentage points, presented this question less as a matter of policy and more as a question of legal precedent. Rather than ask about birthright citizenship directly, the pollster asked the following: “The Supreme Court ruled in 1898 that under the U.S. Constitution anyone born in the United States is a U.S. citizen, regardless of their parents’ citizenship. Do you think the court should keep that ruling in place, or reverse it?” This is, in many ways, fundamentally different than asking if voters support birthright citizenship itself and perhaps contributed to the significant difference in the results.

I did find one survey that showed birthright citizenship underwater by 25 percentage points, conducted by McLaughlin & Associates in May. However, the question wording was so biased that I did not feel comfortable including it on this table. The question first told respondents that “the Fourteenth Amendment was written after the Civil War to make freed slaves citizens. Today it’s interpreted to mean any child born on U.S. soil is automatically a citizen — even if both parents entered illegally or came on a tourist visa.” It then asked if respondents agreed with restricting birthright citizenship to babies born to citizens or those with permanent legal status, asserting this is “what the amendment was actually meant to do.” By presenting this language, it’s essentially giving the answer to respondents rather than asking what they think about the issue.

As we saw with questions about who should be deported in the run-up to the 2024 presidential election, voters have complicated views about who should be granted citizenship when you ask more granular questions.

In the AP-NORC survey, asked generally if all children born in the U.S. should automatically be granted citizenship, 65 percent of respondents said yes. But in the same survey, almost half, 49 percent, said that children born to parents in the country illegally should not be automatically granted citizenship. These views mirror what we saw in advance of the 2024 election, when voters simultaneously said they supported deporting all immigrants in the U.S. illegally but opposed deporting undocumented immigrants who didn’t have a criminal record or separating noncitizen parents from citizen children.

All told Americans seem at least somewhat supportive of birthright citizenship, but they’re also a bit wary of providing citizenship to children of illegal immigrants. How pollsters ask these questions makes a big difference in how Americans answer them, so it can be difficult to get a solid read on what exactly Americans think about the issue.

Share

Other polling nuggets

• Americans’ faith in the success of the negotiations with Iran will lead to lasting peace. Last week, we noted an RMG research survey that showed 47 percent believe the war will end in the near future. This week, in an Ipsos/Reuters poll (June 18 - 22, 2026), just 18 percent said they thought the preliminary agreement was likely to lead to a lasting peace between the two nations, while 63 percent thought it was not likely. A majority — 52 percent — said that considering the costs and benefits, the Iran war was not worth it while 24 percent said it was.

• Every year, the polling company Verasight conducts a survey with questions submitted by attendees at the annual conference for the American Association for Public Opinion Research. This year’s survey was released this week, and it’s full of interesting questions about all kinds of topics including race, AI, and polls themselves. In a question submitted by yours truly, Verasight asked respondents how many Americans they believe approve of the job Trump is doing as president. Our official job approval average for Trump currently shows 37.1 percent of Americans approve of the president; the lowest it has ever been was 36.4 percent in late May. Nevertheless, a majority (52 percent) said they believed less than 35 percent of Americans approve of the president, including 18 percent who said the number was less than 25 percent. Responses, of course, were highly partisan, with 50 percent of Republicans saying they thought over 45 percent of Americans approved of the president, while 73 percent of Democrats said his approval was 35 percent or less.

• According to a new YouGov survey (June 18 - 22, 2026), Americans are pretty confident in our ability to do all kinds of things. Seventy percent thought they could pass the U.S. citizenship test, and majorities thought they could perform CPR or the Heimlich maneuver if needed. On the other hand, just four percent of Americans say they think they could beat a chess grandmaster in a single game of chess. Six percent believe that they could pilot a helicopter to a safe landing or hotwire a car in under a minute. Ten percent said they thought they could navigate to a destination using only the stars. And by a 12 percentage point margin, Americans think they could perform a random stranger’s job for a day, 38 percent to 26 percent (another 36 percent were unsure.)

Polling averages update

All numbers are as of 1:00 PM Eastern on June 26, 2026.

2026 U.S. House generic ballot

The Democratic margin in the 2026 U.S. House generic ballot remained stable this week, rounding to 6 points. And in case you missed it, we made a minor methodological change to how we calculate the average this week.

Trump’s job approval

President Trump’s approval rating among U.S. adults moved about 1 point on net this week, with his approval dropping 0.2 percent and disapproval rising 0.7 percent. An average of 37.1 percent of Americans approve of the president, while an average of 59.4 percent disapprove.

Share

Paid subscribers to 50+1 get access to premium analysis, plus sortable tables and complete data access on our polling website. If you want to follow the 2026 cycle with the best data at your fingertips, become a paid subscriber.

Upgrade to a paid subscription today

Here is what our published generic ballot average looked like yesterday morning, June 24, 2026:

And here is what our generic ballot average looked like this morning, June 25, 2026 (before entering any new polls for the day):

The difference in these trends stems from how the model that powers our averages treats polls of registered voters versus those of likely voters. In this post from the Methods Desk, we explain how these adjustments work and what we changed.

Paid subscribers to 50+1 get access to premium analysis, plus sortable tables and complete data access on our polling website. If you want to follow the 2026 cycle with the best data at your fingertips, become a paid subscriber.

Upgrade your subscription

Buy generic

At early stages of the election cycle, most election polls are conducted among registered voters — who pollsters identify either by using lists of registered voters compiled by states or asking respondents themselves if they are registered to vote — but this group is not a perfect match for the electorate in a given year. That’s because there are many registered voters who won’t actually turn out to vote on Election Day. In 2024, for example, there were roughly 210 million registered voters across the country — but only 152 million cast ballots in the presidential election.

To try to produce poll results that better match election outcomes, pollsters also sometimes sample voters they determine (by various means) to have a high likelihood of actually turning out to vote. The number of likely voter polls usually increases the closer we get to the election.

Historically, likely voter polls have offered more accurate predictions of upcoming elections than polls of registered voters (or those of the entire adult population) — so those are the polls we want to anchor most to when averaging pre-election polls for public consumption. However, there are relatively few of these, and other surveys still provide information our averages can use (mostly in detecting trends within groups over time), so we don’t want to throw them out entirely. How can we make use of polls of the wrong populations when calculating an average among likely voters?

The solution we came up with for our generic ballot average is to estimate the difference in support for each party between polls of both likely and registered voters (and all adults). Then we adjust the polls of the incorrect populations back toward the population of likely voters. We do this by subtracting whatever residual our model detects for each of the two wrong populations (sometimes it’s zero!) from all the polls of that population. This makes polls of registered voters “look like” polls of likely voters. Then, the adjusted polls get run through our average, and ta-da! You have an average of public opinion among likely voters.

For most of the 2026 election cycle, we had been running our average of generic ballot polls anchored to the population of registered voters while we waited for pollsters to release more likely voter samples. But we switched the model to estimate likely voter opinion on June 1, 2026 — resulting in those sharp changes in each party’s support that appeared on the old chart.

Subscribe now

And that brings us to today. We thought those sharp changes were distracting, so we have recalculated the generic ballot average to smooth out the transition from modeling party support among registered voters to likely voters. Because the number of likely voter polls early in a campaign is very small, this leads to an unstable estimate of the gap between likely and registered voters, and a bumpy average early in the cycle when estimating likely voters.

Our solution is straightforward: We calculate both the average of registered and likely voter-adjusted polls and combine them, putting more weight on the likely voter version as the election gets closer. Now, our generic ballot averages start out at the beginning of a cycle adjusting both likely voter and all adult polls toward those of registered voters. At one year before the election, we add in a likely voter-adjusted version, and begin to slowly increase the weight on this average. In March of the election year, it’s 50-50 between the likely and registered voter-adjusted averages, and by June, the average-of-averages is 100% likely voter-adjusted and 0% registered voter-adjusted.

This approach produces our new, smooth curve in the generic ballot average. We are developing a similar technique for our primary and general-election horse-race polling averages.

And here is our latest generic ballot average as of 10:00 AM Eastern on June 25, 2026:

A fully transparent polling aggregator

One of our organizational values here at FiftyPlusOne.news is to be maximally transparent with our methodology and methods-related decision making so users know what they’re getting and why they’re getting it. We believe posts like this are critical for both building in public in general, and for producing the most trusted, most accurate polling averages (and, soon, election forecasts!) available to the public.

If you saw the updated average online before this article was sent out (I made some code changes late at night before the rest of the team was up to copy check this piece), we regret any confusion this caused you.

Paid subscribers to 50+1 get access to premium analysis, plus sortable tables and complete data access on our polling website. If you want to follow the 2026 cycle with the best data at your fingertips, become a paid subscriber.

Upgrade your subscription

Subscribe now

On Monday, June 15, 2026, Iran and the United States announced their agreement to a Memorandum of Understanding in the ongoing conflict. This MOU outlines terms for a potential peace agreement and kicks off a new set of negotiations over the next 60 days to finalize that agreement. Republicans on Capitol Hill have had mixed reactions to the agreement, with some expressing concerns about key elements, such as allowing Iran access to $300 billion in reconstruction funds and the long-term fate of Iran’s military proxies.

Today on The Trendline, we’re taking a look at every poll we’ve seen about Iran since this MOU was signed. Admittedly, that’s not very many polls (just three), but it’s our first glimpse into how Americans feel about the MOU and a possible end to the conflict in Iran.

The first survey comes from RMG Research/Napolitan News Service. It began fielding the same day as the announcement was made (June 15), and was in the field for 2 days. In this poll, we see voters expressing optimism about some elements of the deal. Fifty-eight percent said that the agreement to end the war will be good for the United States, while 20 percent said it would be bad for the U.S. and 21 percent were not sure. Even more voters were optimistic about the deal’s impact on the economy; 67 percent said the agreement would be good for the economy, while just 18 percent said it would be bad and 16 percent were unsure.

However, voters were mixed on whether this deal is actually likely to work. After being told that “the U.S. and Iran have announced an agreement to end the war, though some details will be worked out over the next 60 days,” voters were asked how likely it is that the war will “end in the near future.” Just 13 percent said it was very likely to end in the near future, and 34 percent said somewhat likely. Thirty-two percent said it’s not very likely to end, and 9 percent said it’s not at all likely.

All told, that’s 47 percent who think the war is at least somewhat likely to end soon, and 41 percent who think it’s not. So even though voters think the deal will benefit the U.S., both overall and economically, they’re not quite sure things are going to work out during the negotiation period.

The next survey, from Quantus Insights, was in the field the two days following the announcement of the agreement. In this poll voters were told that “the U.S. and Iran recently announced a preliminary agreement to stop hostilities, reopen shipping through the Strait of Hormuz, and begin 60 days of talks on Iran’s nuclear program and sanctions.” They were then asked if they approve or disapprove of the agreement. Approval was overwhelmingly positive — 56 percent said they approved while just 13 percent said they disapproved. Another 16 percent responded that they “neither approve nor disapprove,” and 15 percent were unsure.

Approval of the deal in the Quantus Insights poll is surprisingly bipartisan. While just 10 percent of Democrats in the survey said they approved of how Trump is handling his job as president, 48 percent said they approved of the Iran MOU while 16 percent disapproved. Among Republicans, 66 percent approved of the MOU, and 55 percent of independents approved.

It’s worth noting, however, that the language of the question may have slightly improved approval ratings for the agreement. The question emphasizes specifically two immediate outcomes: “stopping hostilities” and “reopening shipping through the Strait of Hormuz,” without acknowledging some of the elements in the agreement that have caused concerns, like the reconstruction fund. Admittedly, those more controversial elements are yet to be fully negotiated. While the two items mentioned in the question are expected to occur during the 60-day negotiation period, it may have made the situation sound a bit rosier to poll respondents.

And on the nuclear negotiations, there’s a warning sign in this poll. Voters were told “as part of any deal, the U.S. may push Iran to give up, remove, or destroy its stockpile of highly enriched uranium.” They were then asked whether they would approve of the U.S. “making this a condition of the agreement.” Sixty-three percent said they strongly approved of Iran getting rid of its nuclear material being a condition of the agreement, and another 11 percent said they somewhat approved.

This could be a sticking point in any ultimate negotiations. Last July the U.S. bombed Iran’s nuclear sites, burying the enriched uranium deep underground. According to experts, removal of the nuclear material would be extremely difficult. The MOU itself doesn’t necessarily call for removal of the material but states that it should be disposed of with a “minimum methodology to be down-blending on site.” But even that would first require a risky, complicated extraction process. So if 74 percent of the public thinks this should be a condition of the final agreement, they may find themselves disappointed.

The third and final poll we’ve seen since the MOU was announced was conducted by YouGov on June 18. This poll asked Americans whether, if the final deal that was outlined in the MOU is ultimately signed, the U.S. or Iran will be better off than they were before the start of the conflict. In both questions, a slight plurality of Americans agree that both countries will be better off than they were at the start of the conflict, with 31 percent saying so about America and 36 percent saying so about Iran.

However, it’s clear that Americans think this is a bit better for Iran than for the U.S. Five percentage points more said that Iran will be better off if the deal is signed than said the same about the U.S. And by nine points, more said America will be worse off than said so about Iran.

As with many things about American life, there’s a significant difference by partisanship here. Just 12 percent of Democrats said America would be better off, while 45 percent said it would be worse off. Independents also slightly lean toward negativity about the deal: 24 percent said it would make the U.S. better off, while 31 percent said worse off. However among Republicans there’s much more optimism, with 58 percent saying the deal would make the country better off and just 6 percent saying it’d be worse off.

To sum up where we’re at: Americans are optimistic about some of the terms of the deal and approve of it being signed. They see positive outcomes for the economy in the future as a result of the deal. But they strongly agree that we should require Iran to remove or destroy its nuclear material (which could be impossible),think this is better for Iran than for the U.S. (by a slim margin), and aren’t fully sold on whether or not this deal is even going to work. Of course, we’ll keep an eye on the polls in the coming weeks and bring you more Iran polling as the situation develops.

Share

Other polling nuggets

• According to the latest YouGov/The Economist poll (June 13 - 15, 2026), a majority of Americans support a constitutional amendment to get rid of the Electoral College. Fifty-six percent said they would support an amendment to choose the president by popular vote, while 23 percent were opposed. Even more popular was a potential constitutional amendment to require a balanced federal budget, which Americans favored 72 percent to 7 percent. Some potential amendments didn’t get Americans’ approval, though, such as repealing the Second Amendment (Americans were opposed 55 to 21), allowing presidents to seek more than two terms (Americans were opposed 70 to 16), and eliminating birthright citizenship (Americans were opposed 53 to 29).

• High gas prices continue to cause difficulties for Americans. According to an NPR/PBS News/Marist University poll (June 8 - 11, 2026), 78 percent of Americans said gas prices were putting a strain on their household budgets, with 34 percent saying they’re a major strain and 44 percent a minor strain. The problem appears to be particularly acute for Latinos. While 30 percent of white Americans and 36 percent of Black Americans said gas prices were a major strain, among Latinos that number jumps to a majority of respondents (52 percent). Another 32 percent of Latinos said that gas prices were a minor strain.

• Americans have a complicated view of our shared identity, according to a new NBC News/More Perfect poll conducted by Hart Research Associates/Public Opinion Strategies. The poll found a majority (54 percent) of Americans said that “most Americans share the same core values but disagree about policies and issues,” compared to 44 percent who said that “most Americans have fundamentally different core values.” At the same time, though, just 18 percent said there is more that unites Republicans and Democrats while 80 percent said there is more that divides them. But the division may be perceived to be more at the elite level than anything else: 82 percent agreed that “most ordinary Americans, regardless of political party, have more in common with each other than with people in our country who hold a lot of power.”

• Have you heard The Secret? Apparently, Americans have. A majority of Americans, 53 percent, said they believed that “thinking positively about desired outcomes can directly turn those desires into reality in the physical world” in a YouGov poll (June 12, 2026). Thirty percent said they didn’t believe it. In the same survey, 44 percent said they believe in ghosts, while 42 percent said they didn’t. And 29 percent said they believed that some people can put curses on others, while 57 percent said they didn’t.

Polling averages update

All numbers are as of 1:00 PM Eastern on June 19, 2026.

2026 U.S. House generic ballot

The Democratic margin in the 2026 U.S. House generic ballot rounded to 5 points this week, down from its high of 7 points in late May/early June. About 49 percent of likely voters plan to vote for a Democrat, while 43 percent plan to vote for a Republican.

Trump’s job approval

President Trump’s approval rating among U.S. adults has recovered from its lows, improving by about 2 points on net since last week. An average of 37.3 percent of Americans approve of the president, up from 36.6 last week, while disapproval has fallen from 59.6 percent to 58.7 percent.

Paid subscribers to 50+1 get access to premium analysis, plus sortable tables and complete data access on our polling website. If you want to follow the 2026 cycle with the best data at your fingertips, become a paid subscriber.

Subscribe now

We’re about halfway through the primary calendar for 2026 (27 states have held their primaries thus far), and a curious pattern has emerged. Twenty-nine sitting members of the House of Representatives are running for other offices — a historically high number — but of the 18 that have already faced election, just five of them have won nomination. Republican Reps. Ashley Hinson, Andy Barr, and Kevin Hern won the primaries for open U.S. Senate seats in Iowa, Kentucky, and Oklahoma, respectively, while Republican Reps. Mike Collins and Barry Moore won runoffs for Senate seats in Alabama and Georgia. One other member has also advanced to a runoff, while everyone else has either withdrawn before their primary began or lost their primary.

This amounts to a failure rate of about 67 percent for sitting House members seeking another office, a historically abysmal record. From 1990 through 2024, we identified 279 House members who resigned to run for another office. Of these, just 83 failed to earn their party’s nomination, for a failure rate of around 30 percent. No individual cycle since 1990 has seen House members lose in primaries as frequently as 2026; the previous record was 50 percent, achieved in 2020.

It’s no secret that Americans are frustrated with Congress, which means being a member of Congress may be as much of a liability as a selling point for those seeking another office. Congressional approval in most polls is even lower than approval of President Donald Trump. In a poll from RMG Research/Napolitan News Service earlier this month, 68 percent of Americans said they would vote to replace the entire Congress if it was possible to do so. And 54 percent said a randomly chosen group of American citizens would do a better job than the current Congress, while just 19 percent said they wouldn’t do a better job.

However, this public distaste for Congress isn’t all that new — according to YouGov/Economist polling, approval of Congress hasn’t been above water in at least 17 years — so there’s probably something deeper underlying the poor track record from sitting House members running for other offices this year.

Of course, this record can and will shift by the end of the primary season. Several candidates that we anticipate winning their primaries have yet to face the voters, like Republican Rep. Byron Donalds in Florida’s gubernatorial primary and Republican Rep. Tom Tiffany in Wisconsin’s. But neither of these primary contests is particularly competitive, as GOP support has largely coalesced around each of them. Historically, that’s been a more typical path for House members running for other offices, but not this cycle. The number of House members running in competitive primary contests is especially high this year, and seems to be a major factor driving the high rate of primary losses this year.

Share

More House members are running for other offices, even if their path to victory is unclear

This year’s primary season features 29 sitting House members running for another office, a record in the years since 1990. From 1990 through 2024, an average of just 16 House members have run for other offices each cycle, with the previous record being 23 candidates in 2018.

Of course, election watchers have observed that politicians are retiring from Congress at a particularly high rate this year. In fact, one factor contributing to House members seeking other offices is a high number of open Senate seats, as the Senate has more members leaving office this year than it has in over a decade. But while House members seeking Senate seats is nothing new in Congress, leaving the House to run for a different elected office — even if their odds of winning aren’t particularly strong — can also be a way out that allows a candidate to save face with the public, rather than simply retiring.

Just look at Rep. Chip Roy, who lost the Republican primary for Texas attorney general, or Reps. Nancy Mace and Ralph Norman, who both ran and lost in the same Republican primary for South Carolina governor. All three of them had found themselves at odds with President Trump while serving in the House — a factor that certainly complicated their relationships with the rest of their party, and their paths to seeking other offices.

That goes to show that it’s not just that more sitting House members are running for other offices than in previous cycles; they are also running more often than ever in races that are highly competitive. We classified candidates as competitive or noncompetitive in their primary races based on the gap between the two leading candidates in pre-election polling (when available) and primary election results (if the election has already happened). Candidates were classified as competitive in their primaries if they trailed or led in polling within a month of the primary election by less than 15 percent, or if they won or lost the nomination by less than 15 percent (see below for detailed methodology).

Halfway through the primary season, 2026 already features 12 House members who have run in primary contests where they’re in close competition with other candidates, more than in any other year in the last 18 even-year election cycles.

If we look at all House candidates running for other offices this year, we also see a historically high percentage in competitive races. As of June 17, over half of House members running for other offices in 2026 have faced or are facing a competitive race, with 16 of 29 candidates meeting our definition. That’s the first time this has been the case in the period we looked at since 1990; the previous high was in 2010, when 47 percent of sitting House members seeking higher office were in competitive primary races, and the average from 1990 to 2024 was 31 percent. Contributing to the abysmal track record of House candidates at this point in the cycle: Of the 18 candidates who have already had their primary, 12 of these were in competitive races, and most of them lost. (Meanwhile, three more candidates lost or withdrew from races in which they were not competitive.)

Now, House members have always had a pretty spotty record in competitive primaries; in the 87 competitive primaries we looked at that featured a sitting House member between 1990 and 2024, House members won 43 of them and lost 44 of them, almost perfectly tied. But that isn’t what we’re seeing so far this year. Of the 12 competitive races that have already happened, sitting House members have lost nine of them outright, while two went on to win in a competitive runoff and the last, Rep. Julia Letlow, who is running for Senate in Louisiana, will face a runoff later this month. That’s the worst record for House members in competitive races for other offices in any cycle since 1990, when all 3 House members in competitive races lost.

Share

House members are jumping into races that already have strong candidates

Another unusual feature of this slate of competitive primaries is a high number of House members running for other offices against sitting incumbents or running against each other. Four races in our dataset this cycle feature two sitting House members (which guarantees at least four House members will lose), and another four feature a House member challenging an incumbent. Of the 12 House members either running against each other or an incumbent, nine are in competitive races, seven have already lost their primaries, one is facing a runoff, and only Collins has won.

That’s historically rare. In election cycles since 1990, only one pair of House members on average has run against each other for another office; seven out of the 18 cycles we looked at have no House-on-House matchups. And primarying an incumbent is also quite rare; in only four cycles since 1990 has more than one House member challenged a sitting incumbent, most of them recent: 2006, 2020, 2022, and 2026.

Of the four sitting House members challenging an incumbent this year, two are Republicans who challenged senators that had gotten on Trump’s bad side: Republican Rep. Wesley Hunt in Texas and Letlow. Neither of these was able to win their primary outright; Hunt failed to make the primary runoff, and Letlow is facing a primary runoff on June 27.

A third Republican challenging an incumbent governor, Rep. Dusty Johnson of South Dakota, may have thought the incumbent would be weaker because he had been elevated from lieutenant governor rather than directly elected to the position; Johnson failed to make the runoff in a tight four-way contest in which all four candidates received between 21 and 31 percent of the vote. The lone Democrat challenging a Senate incumbent, Rep. Seth Moulton of Massachusetts, is doing so more for political reasons than ideological ones, campaigning on the message that Sen. Ed Markey is too old to continue to serve.

In addition to more candidates running against each other or against incumbents, this cycle features no House members running unopposed for other offices — the only other time that’s happened in the years we looked at was in 2022. Since 1990, on average three House members have run unopposed for other offices each cycle, with the most coming in 1990 (6 ran unopposed), followed by 1994, 2014, and 2024 (5 each).

Members of Congress have long viewed the House as a stepping stone to other offices, and more than ever are leaving the chamber this year and setting their sights elsewhere. But while most are aiming for the usual targets — Senate seats or governor’s mansions — many of them are falling short, failing even to win their own party’s nomination as they wade into crowded primary fields.

Primaries are one of the main ways that parties determine their future direction and leadership, but as both parties undergo generational and ideological changes, an unusually large number of House members are betting their careers on competitive races they might not win. Whether driven by frustration with Congress, intraparty feuds, or ambitions for higher office, so far most of those bets have come up short. The path forward for both parties is still taking shape in this primary season, but in many cases, it doesn’t seem to run through the House.

Paid subscribers to 50+1 get access to premium analysis, plus sortable tables and complete data access on our polling website. If you want to follow the 2026 cycle with the best data at your fingertips, become a paid subscriber.

Subscribe now

Many thanks to Tia Yang for her support making this article possible, and once it was possible, making it better.

Methodology

In cycles where it was available, the House Press Gallery Departing Member List (previously called the “Casualty List”) was used to identify members who resigned to seek other offices. In other cycles, a variety of sources were used to identify House members who sought other offices. As a result, the data may not be fully comprehensive for older cycles; all members who ran for higher office, such as Senate, governor, or president are included here, but some who ran for lower office may not have been identified.

To determine a candidate’s competitiveness in a race, we first considered election results for each race. In a partisan primary, if a candidate either won or lost by less than 15 percentage points, that candidate is considered competitive. In nonpartisan jungle primaries, if a candidate either won or lost by less than 7.5 percentage points when compared only to other candidates of the same party, that candidate is considered competitive. For presidential elections, any candidate receiving at least 200 delegates at the national convention is considered competitive (none of the candidates in this dataset met that threshold).

For any candidate not considered competitive based on election results, or in races where the primary has not yet occurred, we relied on polls to determine competitiveness. Polls were sourced from a variety of aggregators, including FiftyPlusOne, RealClearPolitics, FiveThirtyEight archives, and Wikipedia. If a candidate led or trailed in pre-election polling by less than 15 percentage points in the most recent polling average produced by a reputable aggregator, that candidate is considered competitive. If no such polling average exists, we take the average of the most recent poll from each pollster conducted within the last month of the campaign. If no such polls exist, we relied exclusively on election results to determine competitiveness.

For jungle primaries, both of the top two performing candidates are considered to have won the primary, regardless of whether a runoff was required. A candidate who failed to advance to a runoff but got the most votes out of any member of their party does not count as a primary loser, since in a traditional primary system that candidate would likely have advanced to the general election. These candidates are instead categorized as “best in primary” and counted among primary winners where relevant.

To request access to the data used in this article, please reach out to data@fiftyplusone.news.

Subscribe now

It’s time for one of the most anticipated sporting events globally — the 2026 World Cup! For the first time, the tournament is being hosted by multiple nations, including the United States, alongside Mexico and Canada. Have Americans noticed? Do they care? This week on The Trendline, we dig in to see how Americans feel about the World Cup, soccer, and the U.S.’s role in this year’s contest.

According to an Ipsos poll (May 29 - 31, 2026), most Americans have at least heard about the World Cup. Fourteen percent say they have heard a lot about the tournament, and another 50 percent have heard a little. But in the same survey, just knowing the event is coming doesn’t translate to enthusiasm: just 7 percent say they are very or extremely excited about the World Cup, while 18 percent say they are somewhat excited and 73 percent say they are not too or not at all excited.

Among fans of professional soccer, though, there’s quite a bit more enthusiasm. According to a YouGov/CBS News poll (Jun. 2 - 4, 2026), only 17 percent of Americans overall say that they are excited about the World Cup, but among fans of professional soccer, that number jumps to 52 percent. Another 24 percent of soccer fans say they are “pleased, but not excited.” However, soccer fandom in the survey is a fairly limited population in the U.S.: just 11 percent of respondents say they are “big fans” of professional soccer, 17 percent say they are “casual fans,” and the vast majority, 71 percent, say they are not fans.

And despite the American men’s team’s spotty record in previous World Cups (of the 22 tournaments that have been held so far, the U.S. last made the top 3 teams in the very first tournament in 1930, and failed to qualify or withdrew in half of the contests), Americans have at least some confidence in their team this year. In the Ipsos poll, 47 percent said they didn’t have high expectations for the U.S. in this year’s World Cup, but 49 percent disagreed. And 26 percent said they would be disappointed if the U.S. doesn’t win the tournament (in their predictions for the tournament, not a single writer for ESPN placed the U.S. in the top four).

Politics, however, has intervened in the World Cup, as it often does in American life. Amid increased immigration scrutiny across the country, reports have swirled about fans and players potentially being turned away at airports; at least one referee from Somalia was denied entry to the U.S. for the tournament. Americans don’t feel great about immigration enforcement at the games either (which the Trump administration has downplayed): in a Washington Post/University of Maryland poll (May 14 - 18, 2026), 65 percent of respondents said they opposed ICE agents being present at matches, while just 35 percent supported the enforcement.

And despite some initial questions, the Iranian men’s team is allowed to participate in the contest, a move generally supported by Americans. In the Washington Post/University of Maryland poll, 75 percent said that the Iranian team should be allowed to play. Sixty-nine percent also said that players from Iran that wanted to stay in the U.S. due to fear of persecution back home should be allowed to remain here.

As to attending the games, some upcoming matches still aren’t sold out, so there’s still some opportunity for fans to catch a match in person. However, the price might be a barrier. The cheapest seats for the U.S. home opener against Paraguay today, for example, are over $1,000. Fifty-nine percent of respondents in the Ipsos poll said it is “too expensive for the average American to attend a game,” and just 2 percent said it was affordable for average Americans (the rest were unsure). In a YouGov poll (May 21 - 24, 2026), just 1 percent of respondents said they would be willing to pay over $1,000 to watch a game that includes the U.S. team; 66 percent said they wouldn’t buy a ticket at any price.

But it’s not only World Cup games that Americans aren’t interested in attending. Asked in the YouGov survey about their previous soccer interest, nearly half (49 percent) said they had never attended a soccer match of any kind, not even a youth game. Just 17 percent said they had previously attended a professional soccer match, and 8 percent said they had been to a previous World Cup game.

However, it’s not all bad news for soccer fans. Interest in the World Cup has been steadily growing, according to YouGov. Ten percent of respondents indicated that they watched matches in 2010, 13 percent in 2014, 17 percent in 2018, and 21 percent in 2022. We’ll find out soon if 2026 continues the trend of Americans learning to love the sport that the rest of the world has long embraced.

Share

Other polling nuggets

• While the political universe may be all abuzz about the Maine Senate race, most Americans are not paying attention. In a Reuters/Ipsos poll (Jun. 3 - 8, 2026) taken the week before his primary, 63 percent of Americans say that they have heard nothing at all about Democratic Senate nominee Graham Platner, while 14 percent say they have heard a lot about him and 22 percent say they have heard a little. More Americans have heard about Ken Paxton, the Republican candidate for Senate in Texas, than Platner: 19 percent have heard a lot about Paxton, 27 percent have heard a little, and 52 percent have heard nothing at all.

• Americans are split about congressional redistricting, according to a new Emerson College poll (Jun. 7 - 8, 2026). Thirty-eight percent told the pollster that mid-decade redistricting is generally a good thing for the country, while 46 percent said it’s generally a bad thing. Support is strongly driven by Republicans, 56 percent of whom said that mid-decade redistricting is a good thing, while 29 percent of Democrats and 30 percent of independents agreed. Majorities (51 percent of both) of Democrats and independents said mid-decade redistricting is generally a bad thing for the country.

• In a new annual poll from the Harvard T.H. Chan School of Public Health and the de Beaumont Foundation (Mar. 19 - Apr. 1, 2026), trust in the nation’s public health sector has dramatically fallen in the last year. Just half of Americans say they trust health recommendations from the CDC; 77 percent trusted the agency in the 2025 survey. The change is being driven mostly by Democrats and independents — trust in the CDC among Democrats fell 58 percentage points since last year’s wave, and among independents it fell 30 percentage points. Trust among Republicans was more or less the same in the two surveys. In addition to lower confidence in the CDC, trust in state and local health departments has also fallen since last year, with 66 percent now trusting their state health department and 70 percent their local health department.

• Belief in aliens is growing, according to a new YouGov/CBS News poll (Jun. 2 - 4, 2026). Sixty-three percent of Americans said they believe there is intelligent life on other planets, up from 48 percent in the 1990s and 56 percent in 2017. And 21 percent believe we have already made contact with intelligent life from other planets, while 33 percent say we will never make contact with intelligent aliens. When it comes to our government’s handling of UFOs, Americans are deeply skeptical. A large majority — 84 percent — say the government knows more about the issue than it’s telling the public.

Polling averages update

All numbers are as of 11:00 AM Eastern on June 12, 2026.

2026 U.S. House generic ballot

The Democratic margin in the 2026 U.S. House generic ballot rounded to 6 points this week, staying more or less steady from last week. About 49 percent of likely voters plan to vote for a Democrat, while 43 percent plan to vote for a Republican.

Trump’s job approval

President Trump’s approval rating among U.S. adults has fallen a tad since last week. An average of 36.6 percent approve of the president, down from 36.9 last week, while his disapproval has risen from 59.4 to 59.6.

—

Paid subscribers to 50+1 get access to premium analysis, plus sortable tables and complete data access on our polling website. If you want to follow the 2026 cycle with the best data at your fingertips, become a paid subscriber.

Subscribe now

On Tuesday, June 9, 2026, the Trump-aligned group Citizens for Sanity announced a six-figure ad buy in Texas. The ad in question features an AI-generated facsimile of James Talarico, wearing a dress and singing a parody of the Sound of Music classic “My Favorite Things,” in which he praises gender transitions for minors.

This ad comes on the heels of the California gubernatorial primary, in which candidate Spencer Pratt grew to public prominence in part due to widely shared AI-generated videos featuring himself and his opponents, often in scenes inspired by comic books. While many of these videos were created by fans, rather than the Pratt campaign, Pratt did share many of them on his social media pages. And last year, in the highly anticipated mayoral race in New York City, Andrew Cuomo drew widespread criticism for his AI-created campaign ads featuring Mayor Zohran Mamdani.

Artificial intelligence is changing the game when it comes to political campaigning, and more than half of campaign consultants say they use AI regularly. But campaigns should be careful how they use it: there’s a growing body of evidence to suggest that far from promoting your desired candidate, this kind of campaign advertising might backfire.

Americans don’t trust AI – and states are responding

Americans are deeply skeptical of artificial intelligence, generally. In a Siena University/The New York Times survey (May 11 - 15, 2026), just 16 percent of registered voters said that AI is “mostly good,” while 35 percent said it is “mostly bad,” 45 percent said “neither good nor bad,” and the rest were unsure. Democrats are slightly more skeptical than the public overall, with 14 percent saying mostly good and 42 percent saying mostly bad. Republicans, on the other hand, are a bit less skeptical than Americans generally, with 22 percent saying AI is mostly good, and 27 percent saying mostly bad.

Asked about the impact of AI on society today, just 10 percent said it has a positive impact in a Reuters/Ipsos poll (May 15 - 18, 2026). A global survey from Baringo reported that worldwide, 70 percent of people are uncomfortable with AI-generated media. And when it comes to political campaigns, Americans have deep concerns. According to a Marist University survey (Mar. 2 - 4, 2026), 85 percent of Americans think that AI-generated political content will spread misinformation related to the upcoming midterms.

Against that backdrop, at least 29 states have enacted laws either prohibiting or restricting the use of artificial intelligence in campaign ads, according to the National Conference of State Legislatures. (This, of course, means that 21 states have NOT restricted the use of AI in campaign materials). However, in many cases these restrictions don’t kick in until 60 or 90 days prior to the election, so for now, it’s a free-for-all.

In most states with AI campaign laws, the restrictions only require that ads depicting a candidate doing or saying something they did not do or say need to be labeled as manipulated or AI-generated. Moreover, in much of the legislation, the restrictions only apply if the media being generated is intended to harm or injure a candidate, making legal enforcement tricky. Pratt, for example, generated AI videos for his campaign without labeling them as such, but if they weren’t “intended to harm a candidate or deceive voters into voting for or against a candidate,” then a disclosure is not required under California law. (We have yet to see if Pratt’s use of AI will be challenged by state authorities.)

A majority of campaign consultants are using AI at work

Despite state restrictions and distrust, campaign operatives are still turning to AI. In a survey of campaign consultants conducted by the American Association of Political Consultants this year, 59 percent of consultants said that they used AI at least weekly in their work, including over a third who said they use it daily. However, most are not using the technology to create ads: just 15 percent say that generating creative content has been a valuable use of AI.

Like most Americans, campaign consults have concerns about the ways that AI can mislead voters. Sixty-five percent of campaign consultants said that the use of deep fakes was among their top three concerns about AI in political campaigns, and 60 percent said that the spread of misinformation/disinformation was in their top 3 concerns.

But as with the general population, slight partisan differences are apparent in how campaign consultants think about artificial intelligence. Twenty percent of Republican consultants said that using AI to produce a campaign ad from beginning to end without disclosing that to the public would be an ethical use of AI; just 11 percent of Democratic consultants and 6 percent of nonpartisan consultants said the same. For AI-generated ads with disclosure, the partisan differences are even stronger: 68 percent of GOP-aligned consultants said that would be an ethical use of the technology, compared to 56 percent of Dem-aligned and 41 percent of nonpartisan consultants.

Share

AI-generated campaign ads might backfire with voters

The AAPC also has research suggesting that synthetic campaign ads might ultimately backfire. In a study from the AAPC Foundation, researchers added AI disclaimers to a variety of videos, some real and some fake, to test how voters responded.

When a video included an AI disclaimer, voters were significantly more distrustful of its content, and had increased skepticism about the message. Voters paid closer attention to videos with noticeable disclaimers, but were generally less receptive to the ad’s message when disclaimers were apparent. That is to say, if you put out an AI-generated ad of Talarico singing about how much he loves trans children, what you actually may be doing is making voters MORE skeptical of the claim that Talarico loves trans children. And once that trust has eroded, if another, more credible ad tries to try to make the same claim, it may be harder to convince voters of its veracity.

Incidentally, corporate America seems to have taken heed of the erosion of trust stemming from AI slop videos much more quickly than campaigns. There’s no shortage of articles from corporate consultants about maintaining brand trust in the age of AI. Campaigns, however, especially Republican campaigns, seem to not have gotten the hint yet. While we can’t know for sure why that is, it may be that President Donald Trump’s fondness for the technology (as evidenced by his Truth Social feed) has provided something of a permission structure for Republican candidates to engage in freer use of AI than we see in other areas of public life.

But if these sorts of AI-generated political ads don’t actually serve to improve your preferred candidates’ standing with voters, who are they for? I can only speculate here, but the experience of the Pratt campaign seems to suggest that these ads and videos are generated less to appeal to actual voters, and more to generate viral spread among online influencers that already like and support your candidate. That can lead to improved fundraising, more followers, and potentially a larger national profile.

Crucially, though, a large national profile isn’t necessarily going to help a candidate win a mayoral race in a specific city. What you actually need to win a mayoral race is more votes than the other candidates, and this sort of advertising may be accomplishing the opposite of that goal.