Subscribe now

On Thursday, April 2, 2026, President Donald Trump fired Attorney General Pam Bondi, a move that had been rumored to be in the works for some time. Bondi joins former Secretary of Homeland Security Kristi Noem as the second cabinet-level official to lose their job in Trump’s second term, within the space of less than a month. So this week on The Trendline, we’re taking a look back on how Americans have viewed Bondi, her handling of her job, and the Department of Justice under her leadership.

At the time of her appointment, the public seemed more or less okay with the choice. In a YouGov/The Economist survey taken just as Trump was being inaugurated (Jan. 19 - 21, 2025), 40 percent of respondents said they approved of her nomination, while 32 percent disapproved. Of the 9 cabinet-level appointees asked about in the survey, this net approval of 8 points was the second strongest, beaten only by approval of Marco Rubio as nominee for Secretary of State (48 percent approved, 32 percent disapproved).

This initial support came in the context of a wave of media, in which Bondi promised ahead of her confirmation vote that she would not “politicize” the Justice Department, and that justice would be “administered evenhandedly” through the country. Compared to some of Trump’s other picks, such as Secretary of Defense Pete Hegseth or Secretary of Homeland Security Robert F. Kennedy Jr., Bondi was seen as a more conventional choice: a state attorney general being elevated to the federal level.

Bondi was confirmed on Feb. 2, 2025, and quickly stumbled into her first mistakes. On Feb. 21, 2025, she told Fox News host John Roberts that the Epstein “client list” was “sitting on my desk right now to review,” causing months of speculation about what she might mean. Later that month, she released the “Epstein Files Part 1,” largely consisting of material that had previously been made public. By July, the Justice Department had denied Bondi’s claims, releasing a memo saying that there was no evidence Epstein was murdered, or that a “client list” existed.

Bondi did not fare well in the polling after this memo was published. In an AtlasIntel survey at the time (Jul. 13 - 18, 2025), 74 percent of voters said they believed that an authentic client list existed based on Bondy’s February statements, and 54 percent said that the DOJ was engaged in “a cover-up of Trump’s previous relationship with Jeffrey Epstein.” Asked if Bondi should remain Attorney General “after the DOJ’s July 2025 memo contradicted her earlier statements about a Jeffrey Epstein ‘client list’,” 28 percent said they thought she should remain, while 51 percent said she should resign.

And while the scandal of the Epstein investigation has swirled in the background throughout Trump’s second term, it’s hardly the only issue Bondi has had to contend with. Facing pressure from Trump, the Justice Department began several prosecutions and investigations into Trump’s perceived enemies, such as James Comey in September 2025 and Leticia James in October 2025. (Both cases were later dismissed in court.)

These moves also didn’t sit particularly well with the public. In a Marquette Law School poll (Nov. 5 - 12, 2025), 55 percent of respondents said that the Justice Department had “filed unjustified cases against Donald Trump’s political opponents,” while 45 percent said the cases were justified. In the same poll, 58 percent said that cases filed against Trump in 2022-2024 were justified, while 42 percent thought they were unjustified, indicating that the public wasn’t just concerned about cases against political figures, but the Trump Justice Department in particular. And in a survey from YouGov/The Economist (Sept. 26 - 29, 2025), 55 percent of respondents said they thought Trump was “directing the Justice Department to go after his political enemies,” while just 25 percent said they thought he wasn’t.

This sentiment has persisted over time: in a YouGov/The Economist poll taken this month (Mar. 13 - 16, 2026), the same question was asked. In that poll, 56 percent of respondents said they thought Trump was using the Justice Department to go after enemies, while 23 percent disagreed. And the Epstein controversy has continued to swirl, as more files have been released and, controversially, some have not. Voters continue to be skeptical of the administration on the issue. In a University of Massachusetts/YouGov survey taken last week (Mar. 20 - 25, 2026), 59 percent of respondents agreed that the Trump administration is “hiding important information about the Epstein case,” while 18 percent disagreed and 24 percent said they “neither agree nor disagree.”

Throughout each of these scandals, Bondi’s unpopularity has continued to grow. While favorability or approval for cabinet level officials is not asked in every survey, we compiled all the instances where YouGov/The Economist asked respondents about Bondi’s favorability.

While her favorable rating has barely changed over time (it has been between 22 and 27 in every instance), her unfavorable rating has slowing grown, from 24 percent in late January to 48 percent in the most recent survey, conducted March 6 - 9, 2026. And in that survey, Americans also said that they supported Bondi’s firing more than any other Cabinet secretary, both in raw numbers and on net.

In all, Bondi served through a tumultuous time in the DOJ. Americans’ initial support and optimism of her nomination as a potentially more conventional Cabinet member was challenged almost immediately by her handling of the Epstein investigation, and her continued perceived missteps only served to increase her unpopularity. Now, we wait to see who Trump will nominate to try to pick up the pieces.

Share

Other polling nuggets

• If you think modern smartphones have gotten too big, you are not alone. According to a new Ipsos survey (Mar. 24 - 25, 2026), eight in ten adults say they prefer a smartphone they can operate with one hand, and the same share say it’s essential that their phone fit into a standard pocket without sticking out or feeling bulky. A majority (51 percent) say modern smartphones have become too wide to grip securely and comfortably, and nearly half (49 percent) wish there were more smaller options on the market. Fewer than half (48 percent) say they’d be willing to accept a bigger, heavier device in exchange for more screen space — suggesting that for most people, the tradeoff just isn’t worth it.

• If you could go to the moon — and were guaranteed a safe return — would you? According to a new YouGov survey (Apr. 1, 2026), a slim plurality of Americans say yes: 44 percent said they’d want to make the trip, while 40 percent said they would not (15 percent weren’t sure). Younger Americans were more likely to say they would take the trip, with 53 percent of those under 31 saying they’d go to the moon compared to just 31 percent among seniors. Women also took the less risky approach, saying 50 to 33 percent they would not take the trip, compared to 56 percent of men who said they would.

• A new Washington Post/George Mason University poll of Virginia’s redistricting referendum finds a tight race ahead of the April 21 election. In a poll that has Trump’s approval rating at 40% among registered voters, the referendum is ahead by just 6 points, 53 to 47 percent. The narrow margin may come down to turnout: 85 percent of Republican and Republican-leaning independents said they are certain to vote or already have, compared to 77 percent of Democrats. Early voting data has reinforced those concerns, with ballots in Republican-leaning districts outpacing those in Democratic-leaning ones.

Subscribe now

Polling averages update

All numbers are as of 11:00 AM Eastern on April 3, 2026.

2026 U.S. House generic ballot

Democrats have held their lead on the 2026 U.S. House generic ballot at +5 (after rounding) over the last week, with 47 percent of registered voters saying they’ll cast their ballot for Democrats this fall versus 42 percent for the Republicans.

Trump’s job approval

Another week, another new low in President Trump’s approval rating. He fell to 47.1 percent approval as of Friday morning, a new low for his second term and rivaling his lowest point ever after the Jan. 6 riot at the U.S. capitol. The percent of Americans who disapprove of the job he is doing also rose from 57.8 to 58.8 percent. The net rating of -21.7 is also a new low.

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

Subscribe now

On February 14, 2026, the U.S. Senate failed to reach an agreement on funding for the Department of Homeland Security, the federal department that houses agencies such as Immigration and Customs Enforcement, the Coast Guard, and the Transportation Security Administration. This triggered a partial government shutdown affecting only the DHS. Democrats insisted that funding not be provided to ICE until changes are made to the way they enforce immigration laws, and Republicans insisted that the entire department be funded together. (Early Friday morning, March 27, Republicans in the Senate gave in and voted in favor of the Democratic plan to fund all offices in DHS except for ICE. The bill now heads to the House of Representatives.)

In the early days of the shutdown, barely anyone noticed (or cared). In an RMG Research/Napolitan News Service poll (Feb. 23 - 24, 2026) conducted 10 days after the shutdown began, just 57 percent of registered voters knew that there was a partial government shutdown ongoing. And only 4 percent of respondents said that the shutdown had “a lot” of impact on their daily life.

For comparison, that is far less than people knew about the shutdown last fall. In a poll taken at a similar point in the life of the fall shutdown (Oct. 13 - 14, 2025), 74 percent knew there was a government shutdown. During the fall shutdown, voter awareness grew over time: In the RMG Research/Napolitan News Service polling, by early November last year, 85 percent of voters knew there was a shutdown.

But this time around, it’s not clear that voters have tuned in in the same way. In a YouGov/CBS News poll taken last week (Mar. 17 - 20, 2026), just 55 percent said they were following news about the shutdown very or somewhat closely. And a third of respondents said they weren’t sure of the Democrats’ or Republicans’ positions on the issue (in reality, this is probably an underestimate, since some respondents might reflexively defend their “side” rather than admit they don’t know what’s going on).

Even so, in the early days of the current shutdown, it looked like the Democrats had a decent level of support for their position. In a Navigator Research poll (Feb. 19 - 23, 2026), 58 percent of voters said they agreed more that congress should “withhold funding from ICE until it changes its approach, even if that means shutting down the Department of Homeland Security,” while 34 percent agreed that they should “continue to fund ICE in its current form to reopen the Department of Homeland Security.”

However, as TSA agents have gone weeks without a paycheck, Americans are starting to see chaos at the nation’s largest airports. Hundreds of TSA agents have quit, and some airports have reported call-out rates among agents of up to 40 percent. And while Americans may not be paying close attention to the shutdown, they certainly noticed the consequences: in a YouGov poll this week (Mar. 25, 2026), 65 percent of Americans said that security wait times at airports were longer than average or among the longest they’ve ever been. Just 10 percent said they were average or shorter than usual. When asked how much the shutdown had increased wait times, 57 percent said “a great deal,” and another 16 percent said “a moderate amount.”

This increased attention on airports and TSA may have weakened Democrats’ hand on the issue. While the February polling from Navigator Research showed Democrats’ position ahead by 24 points, in a mid-March survey (Mar. 12 - 16, 2026), that support had cooled. Using slightly different question wording than the February survey, the March poll found that 49 percent of registered voters preferred Congress to “withhold all funding from the Department of Homeland Security, including funding for the TSA (Transportation Security Administration), until ICE changes its approach,” while 36 percent preferred Congress to fully fund the department, a margin of just 13 points.

As the chaos increased, several approaches to resolve the situation were proposed. Elon Musk offered to pay TSA agents temporarily. Senators offered a compromise deal that would fund parts of DHS, including TSA, but would not include funding for ICE and some other immigration-related departments. Earlier this week, President Donald Trump rejected both of these proposals, instead sending ICE agents to support the TSA in some airports. According to a Verasight/Strength in Numbers poll (Mar. 24 - 25, 2026), both these moves were underwater: Americans disapproved of deploying ICE to airports by 13 percentage points, and disapproved of Trump rejecting the Senate compromise by 31 percentage points.

Finally, in the wee small hours of Friday morning, the Senate went ahead and passed their compromise funding deal, despite the president saying he didn’t support it. And Trump, perhaps realizing how much stress this was causing to American travelers (63 percent of Americans told YouGov that disruptions and delays at airports are very or somewhat stressful), also found a workaround: on Thursday evening, he declared a national emergency, opening a pathway for him to spend funds on pay for TSA agents without having to wait for Congress to allocate them.

It’s yet to be seen whether these moves will help alleviate the travel stress plaguing Americans. We don’t know how quickly DHS will be able to issue backpay for agents, and with hundreds having already quit their jobs, there may still be disruption at some airports. However, it’s quite possible that if the issues with the TSA are more or less resolved via Trump’s emergency declaration, this shutdown will fade back into the background again. That could lead it to drag on even longer if Trump vetoes the legislative solution Senators have crafted, or the House fails to pass it. The shutdown last fall was 43 days, the longest in American history; today is day 41 of the DHS shutdown, so we could be poised to make history again.

Share

Other polling nuggets

• Americans’ top foreign policy priority related to the war in Iran is keeping gas prices low, a new poll from the Associated Press and NORC finds (Mar. 19 - 23, 2026). The survey asked U.S. adults how important it was for the U.S. to pursue four goals in foreign policy: preventing gas prices from rising, preventing Iran from obtaining a nuclear weapon, preventing Iran from threatening Israel, and replacing Iran’s government with one that is “friendlier to U.S. interests.” In total, 67 percent of adults said it was extremely or very important to keep gas prices low, whereas 65 percent said the U.S. needed to keep Iran from getting a nuclear weapon and 39 percent said it was important to protect Israel. Regime change was the least popular goal; just 33 percent of Americans said it was extremely or very important to replace Iran’s government with a more friendly alternative.

• A new Fox News poll (Mar. 20 - 23, 2026) finds half of voters say they are falling behind in daily economic life, and 7-in-10 say neither party has a plan to bring prices down The 46 percent who say they are personally falling behind financially is near the survey’s all-time record — and 61 percent say they wouldn’t be able to pay their bills if they missed more than two paychecks. That’s up from 54 percent in 2023.

• For the first time in Gallup’s post-pandemic tracking, U.S. workers are more likely to say they are personally struggling in their lives (49 percent) versus thriving (46 percent). Just 28 percent of workers say now is a good time to find a quality job in the U.S., down from 70 percent in the middle of 2022. Job satisfaction has dropped across the board, but especially for federal workers: their happiness is down 12 points, double the decline among other workers.

• Americans think artificial intelligence would probably beat them at picking stocks, but are far less confident in AI’s ability to forecast elections. In a YouGov poll (Mar. 25, 2026), 17 percent said they’d beat a chatbot in the stock market, while 29 percent thought the robots would win out. But when it comes to elections, 19 percent thought AI would do better, while 21 percent thought they would beat the chatbots.

Share

Polling averages update

All numbers are as of 11:00 AM Eastern on March 27, 2026.

2026 U.S. House generic ballot

Democrats have held their lead on the 2026 U.S. House generic ballot at +5 (after rounding) over the last week, with 47 percent of registered voters saying they’ll cast their ballot for Democrats this fall versus 42 percent for the Republicans.

Trump’s job approval

Another week, another new low in President Trump’s approval rating. POTUS sagged to 37.9 percent this week, the first time in his second term that his approval percentage started with the number 37. His disapproval rating also rose slightly from 57.2 to 57.8 percent. The net rating of -19.9 is also a new low.

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

Last week, Axios reported on a new maternal-health survey commissioned by the organization Heartland Forward. The poll found that “nearly 9 in 10 Americans (88%) view maternal mortality as a serious problem in the United States. Yet fewer than half (43.6%) understand the U.S. has a higher maternal mortality rate than peer nations.”

Axios’ write-up of the survey looks like any other poll release. The bottom of the “poll” release, however, looks like this:

The survey is not a set of interviews with humans, explains Aaru, the company that conducted the “poll,” but instead based on interviews with artificial “agents” — computer programs that simulate humans by inputting demographic and other data into a series of artificial intelligence and machine learning systems, often referred to as “synthetic data.”

This is not an isolated case. In fact, it’s not even the only case this week! Here’s the methodology for a new “poll” that was released by The Public Sentiment Institute from Monday, March 21, 2026:

The business of synthetic data is booming. Aaru has been pitching synthetic political research for some time; Semafor reported in 2024 on the company’s use of AI agents instead of human respondents in election-related work. Meanwhile, Qualtrics announced this month that clients can now combine synthetic and human panels on one platform. And in February, a company called Simile raised $100 million to build AI systems for predicting human behavior.

I think we are going to see a lot more of these AI-generated polls in the future. And since our job here at FiftyPlusOne.news is to aggregate polls to better estimate issue attitudes and voting behavior, we decided we should have a public policy for dealing with them. In short: 50+1 will not be using any “poll” that includes AI-generated “survey respondents” in our work. Here’s why.

First principles: a poll measures opinion by asking people how they feel

We see two big issues in using AI-generated survey results: one based in theory about what polling is for, and one statistical.

When it comes to the purpose of polling, the core issue with AI-generated data is simple: polls are supposed to be a way of talking to people not predictions of what people would say by algorithmic systems, AI-generated or otherwise.

In 2022 I wrote a book about how polls work and why they are important to the democratic process. A “poll” is something that measures public opinion by interviewing people selected to represent a population. That is the basic logic of survey research, and it remains the standard account of how opinion polling works. When George Gallup was developing some of the foundational ideas about political polling, he intended polls to “take the pulse of democracy.” Surveys, from the democratic point of view (note lower-case “d”), are just another way of talking to the voters. The political scientist Sidney Verba used to say that polls do what democracy is supposed to do in its most ideal form, giving everyone an equal chance to participate in the democratic process.

Synthetic systems do something different. They do not measure opinion by asking people what they think. They infer what people would probably say from demographic profiles, training data, prior surveys, text corpora, behavioral signals, or other model inputs. That sounds fancy, but primarily the way this happens is a researcher tells a large language model to act like a person given some set of demographic variables. What would you say in response to X question if you were a white, 30-44 year old man without a college education that lived in Virginia and made between 50,000 and 100,000 dollars per year? What if you were afraid of heights?

To be clear, such simulations are useful for many purposes. They may even produce estimates that are close to observed public opinion on one or many questions. I have written several academic reports on these approaches so I may have a better idea than most people about what’s going on under the hood here, and I think there is potential for some very accurate emulation of human behavior in the future — perhaps even the near future.

But the predictions these simulations spit out are still only model outputs, not direct measurements of how people feel. And that is why we do not consider such predictions to constitute “polls.”

Why that distinction matters for us at 50+1

The second concern we have is that aggregating predictions from outside models could violate the underlying statistical properties of polling aggregation and election forecasting.

See, polling averages are not just piles of numbers. They work because real polls share certain statistical properties, and can be “combined” using statistical methodologies that detect trends in noisy data. The most fundamental property of these models is that each survey is based on a finite set of actual respondents — and so comes with a measurable level of uncertainty to communicate to the model. On top of that, different pollsters make different choices about sampling, fieldwork, weighting, mode, screening, and question wording that can cause additional bias and error. Averaging helps reduce some noise between pollsters because the mistakes are not perfectly shared across firms.

Synthetic outputs, however, do not have the same type of error structure. If no humans were interviewed, there is no respondent-level sampling error in the usual sense. You can generate 500 synthetic respondents or 50,000, but that does not create new information about what the public thinks. It just produces more draws from the same underlying model. Evidence suggests simulated respondents tend to cluster around an answer more than human respondents, creating both high bias and low variance. (This is not something you want in a polling average.)

Just as important, synthetic estimates are likely to be highly correlated with one another — and in some cases with the polling ecosystem more broadly — because they often rely on overlapping sources of information. They may draw on similar priors, similar public text, similar demographic assumptions, similar media signals, and sometimes even prior polls themselves. If synthetic data is based on polls, and polls are based on synthetic data, then we enter a feedback loop where actual public opinion ends up lost.

That means adding synthetic outputs to a polling average would not diversify error in the way that combining independent surveys can. It would risk hard-coding model-based bias into the average. And our models wouldn’t know how much to update underlying trends in opinion given this new data, because they don’t know how much information these new data points contain.

You don’t have to take my word for it. In their 2024 Political Analysis paper, Synthetic Replacements for Human Survey Data? The Perils of Large Language Models, Joshua Clinton and Jennifer Larson find that while ChatGPT can sometimes reproduce broad averages, it is “not reliable for statistical inference”: the responses show too little variation relative to real surveys, and the estimated relationships often differ substantively from those found in the American National Election Study, a large, annual gold-standard survey of American voters.

A newer February 2026 paper, Do LLMs Track Public Opinion?, reaches a similarly skeptical conclusion. Using daily LLM queries benchmarked against high-quality 2024 election-cycle polls, the authors find “systematic directional miscalibration.” In their headline result, every model overpredicted Kamala Harris’s favorability — often by large margins.

Our policy

So our policy is straightforward. FiftyPlusOne will not collect or aggregate synthetic data in our polling averages or election forecasts. If a source does not interview real human beings, it is not a poll for our purposes.

In addition, we will not aggregate predictions generated by AI-augmented approaches, such as Aaru’s and others, and those based purely on statistical models. This policy applies to fully synthetic systems and to hybrid designs that blend a human interviews with model-generated respondents.

We care about what Americans think, because our government is intended to serve the people. Until robots get the franchise, it’ll be all human at 50+1.

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 recent weeks, there have been a spate of new concerns about prediction markets, online betting sites where everyday Americans can gamble on pretty much anything. Concerns have been raised about insiders leveraging private information to make money on topics as diverse as the television show Survivor, the outcome of this year’s Academy Awards, and even possibly government officials betting on the war in Iran.

Amid the handwringing about whether the casino-ification of American life might be bad for society, Arizona Attorney General Kris Mayes brought criminal charges against Kalshi, one of the major players in the business, and Democratic politicians introduced legislation to ban bets on government actions, including war, and prevent insiders from trading on events where they either know or control the outcome.

So, what do Americans think about these markets? Who’s using them, and what are they using them for? This week on The Trendline, we take a look at what Americans have told pollsters about these new, often risky, financial markets.

Prediction market uptake approaching sports betting sites among young men

According to a poll conducted by Ipsos on behalf of the American Institute for Boys and Men (Feb. 27 - Mar. 1, 2026), most Americans aren’t all that familiar with prediction markets. Just 21 percent said they were very or somewhat familiar with the platforms, another 36 percent said they had heard of them, and 42 percent said they had not heard about them. When it comes to actually using these platforms, just three percent of respondents said they had used Kalshi, one more regulated online option, in the last six months, and 1 percent said they had used Polymarket, another prediction site that lets users wager with cryptocurrency pseudo-anonymously. To put that in context, according to the survey, 6 percent of respondents said they had used DraftKings in the past 6 months.

However, the Ipsos/AIBM poll also included an oversample of Americans under the age of 35. And among younger users, particularly young men, things look a bit different. Among men aged 18-24, 29 percent said they were familiar with prediction markets, and that jumps to 42 percent among men aged 25-34. Another third of young men said they had heard about these markets. And while just 3 percent of the population overall said they had used Kalshi in the past six months, that number more than doubles among young men: 7 percent of men aged 18-24 said they had used the platform, and 8 percent of men aged 25-34 said the same. Asked about a variety of betting-related apps, young men were much more likely than the population overall to say they had used them.

In another survey conducted by Echelon Insights for Paradigm (Feb. 13 - 18, 2026), 17 percent of voters said they had ever placed a bet on a prediction market. As with the Ipsos/AIBM poll, they find that these bettors tend to be younger, with 38 percent of those aged 18-34 saying they had placed a bet, compared to just 3 percent of those aged 65 and up. Moreover, both surveys show that nonwhite Americans are more likely to use prediction markets. For example, in the Echelon Insights/Paradigm poll, 13 percent of white Americans said they had placed a bet, while 25 percent of nonwhite Americans said the same.

As to what users are betting on, the dominant category is sports. Forty-eight percent of those who told Ipsos/AIBM they had used a prediction market in the last six months said they purchase sports contracts. (The sample size of respondents who told Ipsos/AIBM they used prediction markets in the last six months is 96, so treat this paragraph and the following chart with caution.) Following sports, 38 percent said they used the markets for politics contracts, and 36 percent said they used them for contracts related to financial markets.

The fact that sports contracts are so popular on these platforms has caused some consternation with the states. While the Commodity Futures Trading Commission (CFTC) currently holds authority over prediction markets, maintaining that they sell a narrow financial instrument called an “event” or “futures contract”, the authority to legalize and regulate sports betting resides with the states. Lawmakers in several states have argued that sports contracts on prediction markets amount to an illegal sports book, and a number of states are considering legislation to limit prediction markets’ ability to offer these and other controversial contracts. Others may follow in Arizona’s footsteps and bring lawsuits against the platforms.

Americans say prediction markets are effectively gambling sites

And frankly, it appears that Americans might agree with the states more than the feds in this case. In the Ipsos/AIBM poll, asked if various risky financial moves were more like gambling or investing, a majority (61 percent) of Americans said that event contracts on prediction markets were closer to gambling, while just 8 percent said they were closer to investing and another 18 percent said they were a mix of both. This is closer to what Americans said about more traditional gambling, like slot machines, than other new financial products, like cryptocurrency, and quite different from how respondents saw more traditional risky financial behavior, like retail investors buying individual stocks.

When it comes to legality of these markets, voters are also more skeptical about prediction markets than other forms of gambling. In the Echelon Insights/Paradigm poll, a majority of voters said that the lottery (61 percent) and casinos (57 percent) generally should be legal. And forty-five percent said that sports betting generally should be legal. But for prediction markets, just 35 percent said they should generally be legal.

As to how to regulate these markets, in the Ipsos/AIBM survey, 66 percent of Americans said that “no regulation at all” for prediction markets would be a bad idea, while just six percent said it would be a good idea. Majorities said they would support regulating them “like online/sports gambling (e.g., age 21+, state-level rules)” and “like financial investing/trading (e.g., age 18+, federal-level rules),” so either approach might satisfy the public’s desire to rein in these markets.

Voters are especially skeptical of some of the more controversial markets that have popped up in recent weeks, such as those about the outbreak of war or other classified military operations. In a Data for Progress poll (Mar. 13 - 15, 2026), 59 percent of likely voters said wagers “on potential government actions, ranging from what politicians will say in speeches to whether the U.S. will carry out a military strike” should not be allowed, compared to 28 percent who said they should. And when asked about a variety of contentious contract types, voters were overwhelmingly concerned about people using markets to gamble on political outcomes.

As prediction markets continue to expand into more and more domains of American life, we can expect that lawmakers will start trying to regulate (and tax) these industries. And so far, it looks like Americans will be on their side.

Share

Other polling nuggets

• The war in Iran is apparently contributing to rising economic pessimism. A new Economist/YouGov poll (Mar. 13–16, 2026) finds 59 percent of Americans say the economy is getting worse — the highest share to say so since October 2022, and up 6 points from the prior week, one of the largest one-week jumps recorded since 2017. The rise is driven mostly by Independents (up 10 points, to 66 percent) and Republicans (up 6 points, to 24 percent). The poll also finds a strong correlation with gas prices: among Americans who say gas prices are up a lot, 76 percent say the economy is getting worse — compared to only 39 percent among those who say gas prices are up a little.

• Trust in federal public health institutions has declined sharply since Trump took office, and the shift is being driven primarily by Democrats, according to the latest Axios/Ipsos American Health Index (Mar. 6–9, 2026). 58 percent of Americans now trust the CDC, up slightly from 54 percent in October, but down from 66 percent in December 2024, with the FDA falling similarly to 53 percent, down from 60 percent in December 2024. On vaccines specifically, 73 percent agree that parents should follow the CDC’s recommended child immunization schedules, down from 81 percent in March, with the share who “strongly agree” falling from 51 percent to 36 percent. Democrats are still twice as likely as Republicans to strongly agree (59 percent vs. 28 percent).

• A new NBC News poll (Feb. 27–Mar. 2, 2026) finds Americans’ views of Israel have almost entirely inverted over the past decade. When asked whether their sympathies lie more with Israelis or Palestinians, 40 percent of registered voters side with Israelis and 39 percent with Palestinians — compared to a 45–13 percent split in favor of Israel in 2013. The shift has been especially dramatic among Democrats: the share viewing Israel positively has fallen from 34 percent in 2023 to 13 percent in 2026, while negative views have spiked from 35 percent to 57 percent. Republicans have moved only slightly, with two-thirds still siding with Israelis. The shift is now actively shaping 2026 Democratic primaries, with candidates navigating pressure from both pro-Israel donors and anti-AIPAC activists.

• Finland has once again topped the World Happiness Report — its ninth consecutive year at No. 1 — while the United States has continued to slide, according to a new Gallup report (published Mar. 18, 2026). The U.S. has declined from 17th to 23rd, a shift driven largely by lower life evaluations among young adults. The Nordic nations continue to dominate the top of the rankings, though Costa Rica entered the top five for the first time — the highest ranking ever for a Latin American country. In the U.S., Canada, Australia, and New Zealand, youth happiness has fallen by an average of 0.86 points on the 0-to-10 scale, bucking a global trend in which young people in most regions are happier today than 20 years ago.

Share

Polling averages update

All numbers are as of 11:00 AM Eastern on March 20, 2026.

2026 U.S. House generic ballot

Democrats have held their lead on the 2026 U.S. House generic ballot at +4.5 over the last week, rounding to 47 percent of the vote for Democrats versus 42 percent for the Republicans.

Trump’s job approval

President Trump’s approval rating hit a new low of 38.2 percent this week. His disapproval rating also held steady from 57.3 percent to 57.2 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

A crosstab tucked into a recent Marist University survey caught the attention of some political commentators: President Donald Trump’s approval rating in the poll among white voters without a college degree is underwater by two percentage points, 46 percent to 48 percent. Among white men without a degree, the survey showed Trump underwater by 3 percentage points, 46 percent to 49 percent, and among white women without a college degree, he was dead even, with 47 percent saying they approved and 47 percent saying they disapproved.

That polling tidbit made its way around social media, with some speculating that the poll indicates that Trump’s support among the white working class — a group he won by nearly 30 points in 2024 — may now be underwater. Others argued that the deterioration of support among white noncollege voters might indicate potential catastrophe for the midterms. Of course, it’s worth acknowledging that “white votes without a degree” is not really the same as “white working class” (whatever that even means), but it is the demographic group for which we at least have some polling data to go on.

So, has Trump really dropped this much among white voters who don’t have a degree? And what does that mean for the midterms? Is the Marist poll an outlier, or part of a bigger trend?

Share

Subscribe now

Breaking down Trump’s approval among noncollege whites

In national polling, Trump’s approval has fallen significantly over the course of his presidency. (Data in this article reflects all polls that were available by 5:00 p.m. Eastern on Mar. 17, 2026.)

When it comes to white voters without a college degree, the polls generally agree with Marist that Trump’s approval has fallen with that demographic as well. But they disagree on Trump’s exact level of support. In our average of polls that break out Trump’s approval rating by demographic group, Trump started his term with nearly a 61 percent approval rating among white voters without a degree, while around 35 percent disapproved; we estimate that today, that approval rating has fallen to around 51 percent, with 47 percent disapproving.

That is certainly a big shift. But when compared to the changes among other groups of voters, the magnitude of the shift among this specific demographic isn’t particularly significant. When we look at shifts in the net approval rating (the difference between the number of respondents who say they approve and the number who say they disapprove of the president), both overall and among white voters with and without a degree, we see parallel movement throughout Trump’s first term.

This sort of uniform change in polling is generally fairly common. It’s far more unusual to see a demographic group moving significantly more than the population overall than following the same pattern, and when that happens, we let you know.

Subscribe now

Noncollege whites still prefer Republicans in the generic ballot

The erosion of Trump’s approval rating among white noncollege voters doesn’t seem to be translating to erosion of support on the generic ballot. Over the course of Trump’s second term, our estimates of support on the generic ballot in November among white college educated voters and white voters without a college degree have barely budged.

From the beginning of our tracking today, Democrats’ margin among white voters with a college degree has increased by less than 3 percentage points, and Republicans’ margin among white voters without a college degree has decreased by less than 3 percentage points. We don’t have enough data to compute an average for white voters by education prior to April 21, 2025, so we aren’t able to see exactly where these groups were at the beginning of Trump’s term. But in our tracking, this movement is fairly similar to what we see on the topline: on April 21, 2025, Democrats were ahead by 2 points in our topline, and are ahead by 5 points today.

Indeed, even in the same Marist poll that showed Trump underwater with white voters without a college degree, Republicans led among the group on the generic ballot. Marist showed Republicans ahead by 17 points with white voters without a degree on the generic ballot, consistent with our estimates. Among white men without a degree, Republicans led by 15 points.

To be sure, there are reasons to be a bit skeptical about the polling among white voters by education. We only have 23 surveys in our dataset that provided crosstabs for this group, which is a fairly small sample size to go on. But if we consider white voters overall, we have 249 surveys, and our estimates for that group show similar movement–white voters have moved about 3 points, on the margin, toward Democrats since April 21, 2025.

And generally speaking, when we break out different demographic groups, we see parallel movements on the generic ballot, suggesting in most cases a fairly uniform swing. There are, of course, exceptions to that rule—particularly when we consider crosstabs by age—but most groups appear to be shifting by about the same amount over time. So that should give us some confidence that the estimate, while based on somewhat thin data, is capturing a realistic snapshot of these demographics.

Regardless of whether white voters without a degree have been moving towards Democrats over the past year, though, analysis from the Pew Research Center suggests that Republicans’ standing with this demographic has been eroding throughout Trump’s time on the national stage. Pew estimates that in 2016, Trump won white voters without a degree by 36 points; by 2020, that margin had shrunk to 32 points, and in 2024, they estimate he won white noncollege voters by 29 points, 64 percent to 35 percent. Trump’s improvement from an 8-point win among all voters without a degree in 2020 to a 14-point win in 2024 appears to be driven entirely by nonwhite voters: Black voters without a degree moved 18 points toward Trump, and Hispanic voters without a degree moved 20 points, per Pew’s estimates.

In some ways, Pew’s numbers jibe with our current polling average: we estimate that around 36 percent of white voters without a degree would vote for a Democrat in November, fairly similar to the 35 percent that Pew estimates Harris won in 2024. It’s the Republican estimate that is dragging down the margin for the GOP here–we estimate that around 54 percent of white noncollege voters would choose a Republican this fall, far less than the 64 percent that voted for Trump in 2024. It may be the case that the roughly 10 percent undecided voters among this demographic will ultimately lean toward Republicans; based on the data we have right now, it’s impossible to say what that group of undecided voters look like.

Here’s the bottom line: Yes, Trump’s approval rating among white voters without a degree has dipped over the last year—but so has his rating among pretty much every demographic group in the country. There is little evidence that this group is moving faster against the president than any other group, and surveys showing Trump underwater with noncollege whites are, for now, outliers. The group is also roughly as GOP-leaning on the House generic ballot as it was at the start of Trump’s presidency — though, even then, polls showed the group moving left compared to where they were in 2024. The data so far suggest the group is not turning on the Republican Party, even if many have soured on the president.

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

Methodology

To estimate demographic subgroup trends for presidential approval and generic ballot, we use an exponentially weighted moving average (EWMA) that adjusts polls for their sample size, recency, subgroup noise, and movement in the Trump’s national approval trend and generic ballot trend since each poll was conducted. The following methodology statement refers to Trump approval polling; the same steps are taken to calculate generic ballot estimates.

We calculate this average in five steps. For any given day, we look at the polls conducted for a subgroup on or before that day, and do the following:

1. First, each poll’s approve and disapprove vote shares are de-trended by subtracting the corresponding national topline estimate on the poll’s end date, isolating each subgroup’s deviation (aka residual) from the national environment on that day. If the national average for Trump’s approval is 45% on the day a poll is released and the subgroup result is 50%, for example, we record the subgroup residual as +5.

2. Second, we calculate a EWMA with a decay rate of 0.96 (roughly a 17-day half-life) on these de-trended residauls from all polls for this subgroup. Then we add back the national topline from whatever day we are running the average to produce trend-adjusted subgroup estimates. This ensures our average accounts both for movement in the national average (which gets polled more often) and a subgroup’s overall lean toward one party.

3. Third, we compute house effects for each pollster in each subgroup based on how much their polls differ from the average. Each firm’s house effect is equal to the average difference between their results and the trendline-adjusted national average on the day the pollster released their results. We then shrink the house effect toward zero to account for noisy residuals, via the formula `adjusted = raw_effect * n / (n + 5)`, where n = the number of polls from a pollster among that subgroup. So a pollster with 1 poll keeps 1/6 (~17%) of their estimated bias. With 5 polls it’s 50%, with 10 polls it’s 67%, etc. This ensures that when we have relatively few results from a given pollster, we aren’t overconfident in their house effects.

4. Fourth, we subtract each pollster’s house effect from their results, and then calculate a new aggregate estimate using poll readings that are adjusted both for house effects and movement in the national generic ballot average. We do this by repeating steps 1 and 2 above, but using the house-effect-adjusted poll results instead of the raw results. That gives us a final EWMA for approve and disapprove shares in the demographic group!

5. The final step is to blend this EWMA with a model-based prior prediction of the average that is used to smooth out phantom swings that can result from noisy data. The modeled prior for each subgroup is simply the national topline plus that group’s average historical residual values — giving us a stable estimate of where the group “should” be on any given day based on the overall national environment (i.e., not one that is not weighted by recency, which can exacerbate noise in the trend).
   
   The amount of weight assigned to the modeled prior vs EWMA depends on the cumulative information provided by the polls for each subgroup: groups with many recent, large-sample polls that consistently agree with each other will be driven almost entirely by the EWMA, while groups with few, small, or noisy polls will lean heavily on the modeled prior. This effective weight also incorporates a reliability score for each subgroup, estimated by comparing the variance of that group’s de-trended poll results to the median variance across all groups. Subgroups whose polls are noisy and inconsistent accumulate effective weight more slowly, keeping them closer to the modeled prior even when poll counts are similar to more stable groups.

The final averages we report for each group is thus equal to a weighted average of the EWMA for that group and the modeled estimate, using the following formulas:

• w = eff_weight / (eff_weight + 20)

• final = w * EWMA + (1 - w) * model_prior

Where eff_weight is the cumulative sum of decayed poll weights (each poll’s sqrt(sample_size) * reliability, decayed by 0.96 per day since the poll was conducted). So at eff_weight = 20, it’s a 50/50 blend. At eff_weight = 60, it’s 75% EWMA. A brand new group with one small poll might have eff_weight around 5, giving ~20% EWMA and ~80% prior.

Subscribe now

Pentagon officials told lawmakers this week that the first week of the ongoing U.S. campaign in Iran has cost over $11 billion, and some estimates show that the war will cost taxpayers nearly $1 billion every day. Americans tell pollsters that the cost of living is the country’s most important problem and they don’t see Trump as focusing on the right issues, so this week on The Trendline we’re checking in on how Americans think about the cost of the war and federal spending more generally.

Across the political spectrum, Americans do express concern about the cost of the new war against Iran. A majority of Republican, Democratic, and Independent adults said they were very or somewhat concerned about the financial costs of U.S. military action in Iran in a Reuters/Ipsos survey published this week (conducted Mar. 6 - 9, 2026). But notably, Republicans are far less concerned than any other group, and the number of Republicans who said they were “very concerned” is 50 percentage points smaller than the number of Democrats who said the same.

Contrast that with, for example, the number of respondents who told the same pollster that they were very or somewhat concerned about “risk to the lives of American military personnel,” where we see agreement across the political spectrum, with at least 85 percent of every group expressing concern, and it appears that funding stands out as a point on which the parties disagree.

And this disagreement may not come as a big surprise. For years, polls have shown differences in how the two parties prioritize federal spending, particularly when it comes to spending on the military and national defense. While Americans across the political spectrum support increased spending on universal social safety net programs, like Social Security and Medicare, they differ sharply on spending on the environment, education, national defense, foreign aid, and ICE, according to new YouGov/The Economist data (Mar. 6 - 9, 2026).

When it comes to national defense in particular, YouGov showed 58 percent of Republicans supported increasing spending, while 6 percent supported decreasing spending. On the other hand, only 20 percent of Democrats supported increased spending on defense, while 38 percent said it should be decreased.

And the parties also differ in terms of what spending they would cut to reduce the federal debt. In a poll conducted by Gallup last year (Sept. 2 - 16, 2025), Democrats’ most favored option for reducing federal debt (other than raising taxes) was cutting defense spending, with 63 percent saying they would support the move. But among Republicans, only 13 percent said they would support cuts to defense spending. In contrast, GOP voters favored cutting Medicaid and food assistance programs (75 percent) and increasing tariffs (89 percent) to decrease the deficit.

It’s expected that in the coming weeks, Congress will consider a supplemental spending bill to fund continued military action in the Middle East. And despite the fact that voters across the political spectrum have been concerned about the rising national debt for years, Republicans appear ready to support it. Democrats, for their part, have not committed to attempting to block the package, so regardless of American’s feelings about defense spending, it may be inevitable that the money gets approved.

Share

New polling average for Illinois Senate primary

This week, 50+1 launched a polling average for the upcoming Democratic primary for U.S. Senator from Illinois (the seat currently held by retiring Sen. Dick Durbin). According to the polls, the race is most likely to be won by Raja Krishnamoorthi, the current U.S. Rep. from Illinois’ Eight Congressional District; Juliana Stratton, the state’s current Lieutenant Governor; or Robin Kelly, the Rep. from IL-02.

We know it’s a little close to the contest to be launching a polling average (voters cast their ballots next week), but we here at 50+1 are still testing out our infrastructure for more primary averages. We expect Illinois will be our final test, and other averages can come along sooner!

As with the Texas elections last week, primary polling averages should be considered noisy and treated with uncertainty. The average poll for the Republican primary for Texas Senator, for example, was off by 7 points in terms of the final vote margin between John Cornyn and Ken Paxton.

Other polling nuggets

• YouGov conducted an interesting survey experiment testing out the public’s approval of Donald Trump firing Secretary of Homeland Security Kristi Noem last week. The experiment asks roughly 550 voters each one of two questions: either (A) “ Do you approve or disapprove of President Trump’s decision to fire Kristi Noem as Secretary of the Department of Homeland Security?“ or (B) “Do you approve or disapprove of the decision to fire Kristi Noem as Secretary of the Department of Homeland Security?”. YouGov found the question wording mentioning Trump received more support, 60% in favor of versus 11% opposed to the firing, compared to the neutral framing, 53% in favor to 11% opposed. The effect is driven mostly by Republicans, who approve more of Noem’s firing when Trump is mentioned.

• A new national survey by the right-leaning pollster Cygnal finds a majority of Americans (55%) believe AI will have a net negative impact on society, but 67% say it’s still important that the U.S. leads in AI development. Young people (ages 18-29) are the most likely to be pessimistic about AI.

• A new Marist University poll reveals voters are anxious about the security of the 2026 election. Just 66% of voters say they are confident that their state or local government “will run a fair and accurate election this November.” Confidence in election administration has fallen particularly with Democrats, Marist notes: In October of 2024, 88% of Democrats told the pollster they were confident in the upcoming election, compared to 72% now. The poll finds 58% of Americans think voters will be turned away at the polls for being ineligible to vote (the poll did not specify whether this turning away would be correct or incorrect; some Americans are, after all, ineligible to vote).

• The 2026 Academy Awards (Oscars) will be held Sunday, March 15, 2026. YouGov asked voters if they’ll be watching, and who their pick for best picture is. Sinners is a crowd favorite with 19% saying it deserves the award. But the popular choice often loses; the same poll found just 2% of adults think last year’s winner, Anora, should have won.

• A new Data for Progress survey finds a slim majority of likely voters — 52% — believe Trump was “at least partly motivated” to take military action against Iran to distract from the Epstein scandal, compared to 40% who say he was not motivated by it at all.
  
  The headline for this survey, sponsored by the digital news publications Zeteo and Drop Site News, was “Majority of Americans Believe Trump Launched Iran War to Cover Up Epstein Scandal.” We think the headline overstates the finding; the question sets a very low bar for agreement, asking voters whether Trump was “at least partly motivated to take military action against Iran in order to distract from the Jeffrey Epstein scandal” or “not motivated at all” by that goal. “At least partly motivated” means even a respondent who attributes minimal weight to the Epstein angle counts as agreeing with the premise.
  
  We think the more electorally significant finding from this poll comes from a separate question altogether. Voters said they would be less likely to support candidates in 2026 and 2028 who support the war in Iran — by a net of around 20 points across three different electoral scenarios, including congressional races, a 2028 presidential candidate, and members who vote for Trump’s Iran supplemental funding request.

Share

Polling averages update

All numbers are as of 11:00 AM Eastern on March 6, 2026..

2026 U.S. House generic ballot

Democrats have held their lead on the 2026 U.S. House generic ballot at +4.7 over the last week, rounding to 47% of the vote for Democrats versus 42% for the Republicans.

Trump’s job approval

President Trump’s approval rating (38.6%) declined within the margin of error from 38.7% in the week ending March 13, 2026. His disapproval rating also “fell” from 57.8% to 57.5, taking him just off an all-time-high net disapproval rating (now -19).

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

Subscribe now

Share

National polling and historical trends

As of 4:00 PM Eastern on Wednesday, March 11, 2026, the FiftyPlusOne generic ballot polling average shows Democrats ahead of Republicans by 4.9 percentage points.

This marks a significant improvement for the Democrats since our tracking began in late January 2025. Then, Republicans were leading in our aggregate by 3.3 percentage points (they won the 2024 U.S. House popular vote by about 2 points). This sort of movement is to be expected; generally speaking, the president’s party tends to lose ground on the generic ballot over time. Between the fall preceding a midterm election and the election itself, the president’s party loses, on average, 3.5 percentage points on the generic ballot (and loses about 6 points if you exclude the 2002 election, for which normal dynamics were disrupted because of the September 11, 2001 terror attacks). If that average were true this year, it suggests that the Democrats should expect to see additional gains between now and November of several points; polling in the fall showed Democrats ahead by around 3 percentage points, so there is still room for further increase.

That said, there are a few exceptions to the rule, so the Democrats shouldn’t be too complacent. As mentioned, a post-9/11 rally around the flag effect buoyed the Republican party to victory in the 2002 midterms, and in 1990, when George H.W. Bush was president, the Democrats had an unusually good midterm cycle, holding an early lead in the polls of 10 points, which fell to “just” 8 in November. So while there’s a lot of time between now and November for either party to pick up ground, usually in this situation, the out-party — Democrats — is the one who does.

To better understand where there might be opportunities or pitfalls for either party, we gathered up all the crosstabs to get a feel for how key demographic groups are leaning at this point in the race. (Note: all data referenced in this article is as of 4:00 PM Eastern on Wednesday, March 11, 2026.)

Age

One of the critical demographic groups that propelled President Donald Trump to victory in 2024 was younger voters. According to the Pew Research Center analysis of voters in the 2024 presidential election, Trump earned 39 percent of the vote among voters aged 18-29, the best performance of his three presidential bids.

According to polling this cycle, the Republicans started off in a decent position with young voters. In late January and early February 2025, crosstabs showed a fairly close race among young voters, with a margin of less than 4 points in either direction.

But young voters quickly moved away from Republicans as Trump’s term unfolded; by the end of March 2025, Democrats led among young voters by around 17 points, a lead that has grown to around 22 points today.

Now, some of the big jumps in the polling among young voters may be driven by a relatively small number of polls in the early days of the race. But we do not see a similar bounce in Democratic vote margin among other age cohorts, for which we have the same number of polls. While all age cohorts considered here have moved toward Democrats on the generic ballot over time, the movement among young voters has been by far the most significant.

Now, we do find that among younger demographics, more voters are likely to say they are undecided in polls — but it’s clear from this data that the improvement for Democrats is not just undecided voters getting off the fence. The Republican share of the vote among young voters has fallen from the low 40s at the beginning of the trend (consistent with Trump’s 2024 performance) to the low 30s today (more in line with the 2020 and 2018 elections). And the number of undecided 18-to-29-year-olds hasn’t meaningfully changed; around 14 percent did not choose a candidate in the early part of the trend, whereas today it’s around 12 percent. So we can be reasonably sure Republicans’ weakened position is not just from voters refusing to tell pollsters how they feel, but them actually changing their minds over the course of Trump’s presidency.1

However, those undecided voters could be hiding a potential upside for Republicans. In the latest Harvard Youth Poll conducted in Fall 2025, 22 percent of 18-29 year old voters that voted for Trump in 2024 were undecided about their midterm vote choice, while just 11 percent of Harris voters said the same. And young people that did not have a college degree (and weren’t pursuing one) were also the most likely group of young voters to say they were undecided: 41 percent of non-college young people said they were undecided, compared to 27 percent of college students and 24 percent of college graduates. According to Pew, Trump won voters without a college degree by 14 points in 2024 — so there may be room for Republicans to improve among younger voters.

Race

When it comes to race, there’s some good news and bad news for both parties.

As with all demographic groups, there has been movement over time towards Democrats among all racial cohorts. And unlike among age cohorts, that movement has been fairly uniform. White, Black, and Hispanic voters, for example, have all shifted 6-7 points towards Democrats, on the margin, since the trend began.2

However, there’s some warning signs for Democrats under the hood. While Black voters have shifted towards Democrats since the trend began, our polling average shows 16 percent of Black voters still plan to vote for the Republican candidate in their local House seat in November. This is fairly similar to the 15 percent of Black voters that Pew estimates voted for Trump in 2024, nearly double his vote share among Black voters in 2020. So even as Democrats are making improvements overall, there may still be some work for them to win back support among this crucial demographic.

On the other hand, Republicans seem to have lost ground compared to 2024 among Hispanic voters.3 Pew estimates that 48 percent of Hispanics voted for Trump in 2024, with him losing the demographic by just 3 percentage points, compared to a loss by 25 percentage points in 2020. But in the latest data, Democrats lead among Hispanic voters by around 16 percentage points. And while 11 percent of Hispanic voters do not indicate they plan to vote for either a Republican or Democrat in November, even if you allocated all those undecided voters to the Republicans (highly unlikely!), the party would still be a point shy of matching Trump’s 2024 performance.

And if you are skeptical of crosstabs as a way to measure movement among demographic groups, elections themselves bear this out. Results of both special and general elections held throughout 2025 and 2026 broadly also suggest Republicans have lost ground with Hispanics relative to Trump’s performance in 2024.

Independents

Among independent voters, we’ve also seen a substantial shift toward Democrats since our generic ballot polling average began. Today, our aggregate of polls shows Democrats winning 42% of independent voters, compared to 28% for Republicans. That compares to 36% for Democrats and 30% for Republicans in January 2025.

However, there’s a significant number of independent voters that do not indicate support for either the Democratic or Republican candidates. Depending on the poll, these voters may be undecided, or may tell pollsters that they don’t plan to vote in November at all. Because few pollsters are currently screening for likely voters, it can be hard to tell how things are really going with the political middle.

To make things even more complicated, different pollsters define “independent” in different ways: some pollsters may push respondents to say whether they lean toward one party or the other, and others may not. How so-called “true” or “pure” independents differ from those that lean toward a party can’t be teased out from the dataset we currently have.

Even given those caveats, it’s hard to see the roughly 14-point lead for Democrats among independents as anything but good news for the party. According to Pew, Trump was dead even with independents in 2024 (their version does not include those that lean toward a party), so that swing probably indicates real movement among these voters.

Geography

When it comes to geography, we see a familiar pattern: urban voters support Democrats, rural voters support Republicans, and suburban voters are somewhere in between.

As with racial demographics, there are some concerning elements for both parties here. First, Republicans are significantly underperforming with rural voters compared to recent elections. As of now, about 38 percent of rural voters plan to vote for a Democrat in November, per our estimates. This is better than any presidential candidate has done with rural voters in at least a decade: according to Pew, Harris earned just 29 percent among rural voters in 2024, and both Biden and Clinton had 34 percent vote share with rural voters. And while improvements among rural voters may not help Democrats that much in the House, where gerrymandering ensures more favorable demographics for both parties in individual races, several potentially competitive Senate seats are in states with significant rural populations, such as Ohio, Iowa, and Texas.

On the other hand, Republicans seem to be holding on to their 2020 and 2024 gains among urban voters. In 2016, Pew estimated that 24 percent of urban voters chose Trump over Clinton, which grew to 32 percent for Trump in 2020 and 33 percent in 2024. In our polling averages today, 36 percent of urban voters plan to vote for the Republican candidate in November. If this holds to November, Democrats could be facing a challenge in key Senate races in states with substantial urban populations, such as Texas.

Overall, the generic ballot polling so far shows Democrats on their way to a decent showing in November. And under the hood, they’ve improved their margins with every demographic group we looked at (even the ones not discussed here) since the beginning of 2025. However, Republicans should be feeling good about retaining some of their recent gains among traditionally Democratic groups, like urban voters and Black voters, which may help offset losses among other demographics.

Democrats are up 5 points overall, per our generic ballot average today. The trend most responsible for this swing (7 points relative to the 2024 election) is the party’s strong showing among political independents (+14 at time of publication). If that holds, the party should win the majority of seats in the U.S. House of Representatives quite easily. But we just have polls, not a crystal ball — so that “if” is still just that, an “if.”

Subscribe now

Share

Methodology

To estimate demographic subgroup trends for the generic congressional ballot, we use an exponentially weighted moving average (EWMA) that adjusts polls for their sample size, recency, subgroup noise, and movement in the national generic ballot since each poll was conducted.

We calculate this average in five steps. For any given day, we look at the polls conducted for a subgroup on or before that day, and do the following:

1. First, each poll’s Democratic and Republican vote shares are de-trended by subtracting the corresponding national topline estimate on the poll’s end date, isolating each subgroup’s deviation (aka residual) from the national environment on that day. If the national average for the Democrats is 45% and the subgroup is 50%, for example, we record the subgroup residual as +5.

2. Second, we calculate a EWMA with a decay rate of 0.96 (roughly a 17-day half-life) on the de-trended values from all polls for this subgroup. Then we add back the national topline to produce trend-adjusted subgroup estimates. This ensures our average accounts both for movement in the national trend and a subgroup’s overall lean toward one party.

3. Third, we compute house effects for each pollster in each subgroup based on how much their polls differ from the average. Each firm’s house effect is equal to the average difference between their results and the trendline-adjusted national average on the day the pollster released their results. We then shrink the house effect toward zero to account for noisy residuals, via the formula `adjusted = raw_effect * n / (n + 5)`, where n = the number of polls from a pollster among that subgroup. So a pollster with 1 poll keeps 1/6 (~17%) of their estimated bias. With 5 polls it’s 50%, with 10 polls it’s 67%, etc. This ensures that when we have relatively few results from a given pollster, we aren’t overconfident in their house effects.

4. Fourth, we subtract each pollster’s house effect from their results, and then calculate a new aggregate estimate that adjusts for house effects and movement in the national generic ballot average. We do this by repeating steps 1 and 2 above, but using the house-effect-adjusted poll results instead of the raw results. That gives us a final EWMA for each party in demographic group!

5. The final step is to blend this EWMA with a model-based prior prediction of each party’s average that is used to smooth out phantom swings in the average that result from noisy data. The modeled prior for each subgroup is simply the national topline plus that group’s average historical offset from the national topline — giving us a stable estimate of where the group “should” be on any given day based on the overall national environment.
   
   The amount of weight assigned to the modeled prior vs EWMA depends on the cumulative information provided by the polls for each subgroup: groups with many recent, large-sample polls that consistently agree with each other will be driven almost entirely by the EWMA, while groups with few, small, or noisy polls will lean heavily on the modeled prior. This effective weight also incorporates a reliability score for each subgroup, estimated by comparing the variance of that group’s de-trended poll results to the median variance across all groups. Subgroups whose polls are noisy and inconsistent accumulate effective weight more slowly, keeping them closer to the modeled prior even when poll counts are similar to more stable groups.

The final average we report for each group is thus equal to an average of the EWMA for that group and the modeled estimate, using the following formulas:

• w = eff_weight / (eff_weight + 20)

• final = w * EWMA + (1 - w) * model_prior`

Where eff_weight is the cumulative sum of decayed poll weights (each poll’s sqrt(sample_size) * reliability, decayed by 0.96 per day since the poll was conducted). So at eff_weight = 20, it’s a 50/50 blend. At eff_weight = 60, it’s 75% EWMA. A brand new group with one small poll might have eff_weight around 5, giving ~20% EWMA and ~80% prior.

After averaging for each group, overlapping age ranges (e.g., 18–24, 25–34, 18–44) are collapsed into four standard buckets (18–29, 30–44, 45–64, 65+) using the approximate share of the population in each group. (While they are not shown here, we do the same thing for education subgroups. Several overlapping categories are collapsed into four tiers based on the proportion of the population that is shared between the original and target groups: high school or less, some college, bachelor’s, or postgraduate. A separate binary college/no-college category is separated out from the more detailed education variables. We will dive into education and race-by-education breaks in a future crosstabs deep dive!)

Footnotes

In the last few weeks, dozens of pollsters have asked Americans their opinions about U.S. military intervention in Iran. Many articles have been written summarizing these polls, so we won’t go into detail rehashing all the results here. The topline takeaway is, generally, Americans disapprove of the United States’ attacks on Iran starting Feb. 28, 2026, with approval roughly in the high 30s/low 40s, while disapproval sits roughly in the high 40s/low 50s.

Instead, this week on The Trendline, we’re taking a slightly different approach. We’re focusing on polling among Republicans, conducted both before and after the strikes, to better understand how the different factions in the Republican party view not just the Iran conflict in particular, but America’s use of military force and foreign policy more broadly. Most of this polling is based on crosstabs from nationwide surveys, as narrow polling among Republicans on niche issues like this is hard to come by, but it still paints a picture of a party that, while they may be on the same sides of most issues, are still divided in many ways.

First, consider Gallup’s latest World Affairs survey (Feb. 2 - 16, 2026) on the priorities of Americans when it comes to foreign policy.

Compared to Americans overall, Republicans are less likely to prioritize altruistic foreign policy goals that explicitly help other countries, such as “defending our allies’ security” and “promoting and defending human rights in other countries.” On the other hand, Republicans are more likely than Americans overall to prioritize foreign policy goals that would explicitly benefit the U.S. or its safety, such as “securing adequate supplies of energy for the U.S.” and “preventing future acts of international terrorism.”

In addition, according to Ipsos/Chicago Council on Global Affairs surveys conducted over the last 15 years, Republicans have always been far more concerned about Iran’s nuclear program than other political groups.

In the most recent Ipsos/Chicago Council survey that asked this question (July 2025), the gap between Republicans’ view of the nuclear program as a “critical threat” and the overall public’s view was the widest the poll has ever measured, at 14 points. And in their latest survey (Feb. 27 - Mar. 1, 2026), conducted just after the latest U.S. campaign began, this issue topped the list of concerns among Republicans, with 88 percent saying they were “very or somewhat concerned” about “the possibility of Iran developing nuclear weapons” (75 percent of the public overall agreed).

Given the high level of concern about Iran’s nuclear program among Republicans, and the general instinct to support actions taken by a president of one’s own party, it may be unsurprising that Republicans are much more supportive of the military action in Iran than other political groups. Indeed, even before the strikes were launched, a YouGov/CBS News poll (Feb. 25 - 27, 2026) found that 84 percent of Republicans said they would favor “the United States taking military action against Iran to try to prevent them from producing nuclear weapons,” while 16 percent opposed. By contrast, among Americans overall, 51 percent said they favored the action and 49 percent were opposed.

But the high support among Republicans may also reflect a tendency for voters to follow the lead of their elected and opinion leaders. After the strikes began, YouGov and CBS News reached back out to the same poll respondents to see how their minds had changed. In a less specific question, respondents were asked if they approved or disapproved of “the U.S. taking military action against Iran.” The public overall had shifted ten points away from supporting the action, with 44 percent approving and 56 percent disapproving. But among Republicans, support for military action had actually increased one point, to 85 percent, with opposition at 15 percent. The write-up of the article mentioned that the “MAGA base”1, in particular, was supportive of the action, but did not provide specific details of how MAGA Republicans and non-MAGA Republicans responded.

Other surveys, however, did provide this breakdown. In a CNN/SSRS survey conducted just after the strikes (Feb. 28 - Mar. 1, 2026), 77 percent of Republicans said they approved of “the U.S. decision to take military action in Iran,” while 23 percent disapproved. But the article noted the following:

MAGA Republicans are 30 points more likely than non-MAGA Republicans to say they strongly approve of the decision to take military action, 34 points likelier to say it will reduce the threat Iran poses to the US and nearly 50 points more likely to say they have a great deal of trust in Trump to make the right decisions about US use of force in Iran.

Now, this may seem counterintuitive to people that equate the MAGA movement with the America First movement, but these two ideas are not quite ideologically aligned. While elites in the party do seem to marry these ideas, a survey from Navigator Research conducted just after the strikes on Iran last summer (June 26 - 30, 2025) shows that is less true among the public at large. The pollsters asked all respondents which of the following two statements they agree with more:

• It’s better for the U.S. when we’re more active and involved in world affairs

• It’s better for the U.S. when we’re less active and involved in world affairs

Respondents that chose the first statement were labeled “interventionist,” and those that chose the second statement were labeled “isolationist.” Among other differences, interventionist Republicans were slightly more likely to identify as MAGA (71 percent) than isolationist Republicans (64 percent). (Interventionist Republicans were also generally much older and more politically engaged than isolationist Republicans.) Among Republicans overall, 50 percent were classified as interventionist, more than among Democrats (48 percent).

Non-MAGA Republicans were also generally far less supportive of Trump than MAGA Republicans. For example, MAGA Republicans were 27 points more likely to say they approved of Trump’s handling of foreign policy than non-MAGA Republicans. On whether Trump made the right decision to “bomb Iran, targeting their nuclear program,” MAGA Republicans were 20 points more likely to say it was the right decision than non-MAGA Republicans. Asked whether the U.S. should continue to use its military in the conflict, 53 percent of Republicans said yes; among MAGA Republicans that number rises to 61 percent. On the other hand, a plurality (46 percent) of non-MAGA Republicans said the U.S. should stop.

And in another Navigator survey, this one conducted just before the most recent military actions in Iran (Feb. 19 - 22, 2026), MAGA Republicans were much more likely to support wars, including regime change wars, than any other political group.

This tendency of MAGA Republicans to prefer interventionist policies is not new or unique to the Iran conflict. In a survey from Angus Reid (Jan. 16 - 20, 2026), respondents were asked whether various actions taken in the first year of the Trump administration made them more pleased or upset. While MAGA Republicans were more pleased than non-MAGA Republicans on every issue tested, the largest gaps were among issues related to unilateral or potentially undemocratic actions by the Trump administration and foreign policy.

In a survey from the same pollster conducted after the strikes (Mar. 2 - 4, 2026), MAGA Republicans were significantly more aggressive than non-MAGA Republicans on the operation. MAGA Republicans, for example, said they would support a U.S. decision to send ground troops to Iran by 49 points, 66 to 17, while non-MAGA Republicans said they would support such a decision by only 2 points, 40 to 38. Asked if the military action in Iran would make the U.S. more or less safe, 76 percent of MAGA Republicans said it would make the country safer, and just 5 percent said it would make the country less safe; among non-MAGA Republicans, just 41 percent thought the conflict would make the country safer, while 21 percent said it would make the country less safe. And MAGA Republicans generally thought the operation would be quicker: 59 percent said they believed the conflict would last up to six months, while just 39 percent of non-MAGA Republicans believed the same.

These polls paint a clear picture: while Republicans are broadly more supportive of military action in Iran than the public at large, the MAGA wing of the party is driving much of that support. MAGA Republicans are more hawkish, more trusting of Trump’s leadership on foreign policy, and more optimistic about the conflict’s outcomes than their non-MAGA counterparts. Far from being the isolationist “America First” bloc they’re often portrayed as, MAGA Republicans consistently emerge in polling as the most interventionist faction in American politics today.

Share

Other polling nuggets

• A new poll from the Vanderbilt Project on Unity and American Democracy (Feb. 20 - 23, 2026) finds broad, bipartisan pessimism about the state of the country. Two-thirds of Americans say the country is on the wrong track, and 73% believe the Founding Fathers would not be proud of the country today. The economic mood is particularly sour: 77% disagree with a question asking if their cost of living has decreased, and 63% rate the current economy as fairly or very bad. The pessimism is especially acute among young people, with just 12%of 18- to 29-year-olds saying they are “extremely proud” to be American, compared to 41% among those 65 and older.

• With the 2026 Winter Olympics just concluded, a new YouGov survey (Feb. 27 - Mar. 2, 2026) finds that Americans broadly like their national sports teams. The U.S. women’s ice hockey team tops the list, with 56% of adults saying they view the team favorably, followed closely by the men’s ice hockey team (54%), the men’s basketball team (53%), women’s basketball (50%), men’s soccer (50%), and the women’s soccer team (48%). Notably, these teams are popular across party lines — the U.S. men’s hockey team, for example, is viewed favorably by 68% of Republicans and 53% of Democrats — making Olympic sports one of the few things Americans can still agree on.

• Interest in weight-loss drugs continues to grow. According to a new Leger survey (Feb. 6 - 9, 2026), 11% of U.S. adults report currently taking a prescription GLP-1 medication like Ozempic or Wegovy, and another 12% of non-users say they’re interested in taking one. Weight loss is the top motivator, cited by 71% of those considering the drugs. Users also report notable lifestyle changes: fewer cravings, smaller restaurant portions, and shifts in spending away from indulgent categories and toward fitness and personal care.

• Trump’s brief improvement among Hispanic voters appears to be fading. According to a new PRRI survey (Feb. 10 - 18, 2026), just 24% of Hispanic Americans now hold a favorable view of the president, down from a high of 39% in September 2024. The decline has been especially sharp among Hispanic Protestants, who dropped from 48% favorability in September 2025 to 37% in February 2026. Overall, just 36% of Americans view Trump favorably, down from 41% in September 2025.

• Support for abolishing Immigration and Customs Enforcement has hit a(nother) new high. According to the latest Economist/YouGov poll (Feb. 27 - Mar. 2, 2026), half of Americans (50%) now support abolishing ICE — the first time the figure has reached the majority threshold in YouGov’s polling. At the same time, opposition has fallen to an all-time low of 39%. For the first time, a majority of independents (52%) back abolishing the agency, and even among Republicans, support has reached a new high of 23%. The shift has coincided with declining trust in ICE overall: 44% of Americans now say they have no confidence in the agency at all, up from 38% in October, and a majority (58%) say ICE uses excessive force.

Share

Polling averages update

All numbers are as of 11:00 AM Eastern on March 6, 2026.

2026 U.S. House generic ballot

Democrats have held their lead on the 2026 U.S. House generic ballot steady over the last week, at about 47% of the vote versus 42% for the Republicans.

Trump’s job approval

After briefly hitting a new net low of -20 in 50+1’s aggregate, Donald Trump’s approval rating reverted back to previous levels somewhat and ends the week at 39% approve and 58% disapprove, among all U.S. adults.

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.

Share

Subscribe now

Footnotes

Subscribe now

By all accounts Peggy Flanagan, Minnesota’s current lieutenant governor, shouldn’t be winning the state’s Democratic primary for this year’s U.S. Senate race. She’s being outraised by a ratio of more than two-to-one. Her main opponent, Rep. Angie Craig, is a well-known House member with a proven track record of winning tough races. And perhaps most importantly, lieutenant governors have an abysmal track record when running for Senate.

A year ago only half of Minnesotans knew who Flanagan was. Yet according to FiftyPlusOne’s collection of polls, Flanagan leads in every single survey pollsters have published in the race so far. Even polls that have been conducted by Flanagan’s opponent have the statewide number two ahead. Our analysis suggests Flanagan is currently defying historical gravity due to a mix of support from statewide leaders and a strong personal brand. A win is not a foregone conclusion, but she has a strong chance of bucking the historical trend.

Lieutenant governors usually lose races for higher office

Flanagan is trying to pull off a feat that is quite rare in recent political history. Since 2010, 10 lieutenant governors have run for Senate without first being appointed to the office. Only one of those succeeded: Pennsylvania’s John Fetterman. And it’s not just the Senate; lieutenant governors have a pretty lousy success rate running for any office. Since 2010, 74 lieutenant governors have run for statewide or federal office, but they’ve won just 36 percent of their races. And that number is perhaps even optimistic for Flanagan, since it includes lieutenant governors who were initially appointed to their seat or who ascended to become governor and then later ran for a full term. These candidates have a leg up on their colleagues, since they can run as an incumbent (although without the full benefits that typical incumbency offers). When these “ascenders” are excluded, the success rate drops to an even lousier 28 percent.

The biggest problem for lieutenant governors might be the job itself. With few exceptions, they just don’t do much. In most states the lieutenant governor is weaker than the governor and lacks a specific domain through which to channel power — secretaries of state, for example, handle elections and business filings; attorneys general get all of the law. The second-in-command usually handles little more than chairing various blue-ribbon commissions or presiding over their state’s Senate (if they even bother to show up, that is). As a result, there aren’t many opportunities to make headlines or raise their profiles in a substantive way that could help them fundraise or win a primary.

That makes for a pretty hefty tide to swim against when running for another office. But Flanagan is staying afloat with ease so far, performing much better than history and the fundamentals would otherwise suggest. What explains her overperformance?

Become a paying subscriber of 50+1 to support independent journalism like this article and get access to complete data tables at our companion polling aggregation website.

Subscribe now

Flanagan has the advantage on local matters

To start with, Flanagan has managed to cultivate more support from local Minnesota politicians than Craig, whose prominent endorsements have largely come from out-of-state officials. Flanagan has the backing of the outgoing senator she’s running to replace, Tina Smith, as well as the state attorney general, state auditor, plus 46 current state legislators. Craig, meanwhile, has endorsements from just 18 sitting state legislators and none from any statewide officeholders.

Flanagan’s high level of support from local politicians looks to have boosted her profile in the state, and given her access to a broader and more lucrative fundraising network, though she still trails Craig by a significant amount. At the end of last year, Craig had raised almost $7 million, and ended 2025 with almost $4 million cash on hand. Comparatively, Flanagan had raised just over $3 million, and had about $800,000 cash on hand. But the momentum behind those donations is key. Craig’s quarterly hauls stayed relatively flat throughout the year, while a string of key endorsements last spring boosted Flanagan to more than double her fundraising haul from the first quarter of 20251, and kept up that pace through the rest of the year.

Flanagan also boasts an impressively popular reputation. In the three public polls of the race that have asked whether voters view the candidates favorably or unfavorably, Flanagan had an average favorable rating of 72 percent among Democratic primary voters. Craig, on the other hand, was viewed favorably by an average of 64 percent of Democratic primary voters in the state. That’s still a respectable level of support, but not the kind of numbers you’d hope to be getting from your own party’s voters as a four-term congresswoman — and one who has already spent over $3 million on the race.

Craig has also stumbled on immigration, which polls show is tied with the cost of living as the top issue in the state today. At the beginning of Trump’s second term, Craig voted for the Laken Riley Act, which requires the Department of Homeland Security to detain unauthorized immigrants who have been accused of certain nonviolent crimes. She also voted for a resolution that expressed “gratitude” to ICE agents, though it also included language that condemned an antisemitic terrorist attack in Colorado.

At the time, some observers thought these votes were politically savvy moves that would counter Republican attacks on Democratic immigration policy made during the 2024 election. But after ICE agents killed two U.S. citizens in Minneapolis, her vote for any Trump-supported immigration measure became an instant liability among Democratic voters, who were flatly opposed to ICE’s actions in Minneapolis and across the country. In a PPP poll for the Flanagan campaign (conducted Jan. 16 - 17, 2026), around two thirds of primary voters said that Craig’s votes for those bills made them less likely to support her.

Craig has said that she regrets voting for the Laken Riley Act, and introduced articles of impeachment against Homeland Security Secretary Kristi Noem following the killing of Renée Good by ICE agents in January. But that hasn’t stopped Flanagan from hammering her on the issue across social media. And in a sign of just how potent immigration enforcement has become in the race, the first thing Sen. Tina Smith mentioned in her video endorsing Flanagan as her successor was ICE’s operations in Minnesota.

Another reason Craig may be struggling to gain traction is that she’s marketing herself as a moderate Democrat, and has drawn little coverage from high-profile endorsers. While Flanagan has rolled out endorsements from progressives like Vermont Sen. Bernie Sanders and Massachusetts Sen. Elizabeth Warren, Craig has emphasized her bipartisan credentials and experience working across the aisle. Yet polling shows that Democrats are looking for their party leaders to fight harder against Republicans. While that’s not an explicitly ideological demand, appeals to bipartisanship and compromise don’t often read as strength. Craig has pivoted recently and wrapped much of her campaign messaging in “fighter” language, but it may be too little too late.

Flanagan certainly isn’t without her own political baggage, though. Last fall, allegations of fraud in the state’s social services system led to Gov. Tim Walz suspending his reelection bid and plenty of criticism by association for Flanagan. She has tried to slough off responsibility and lay the blame at Walz’s feet, but is still facing attacks from the Craig campaign for failing to own up to the fraud that happened partly under her watch.

Share

Win rates vary drastically across cycles

It is possible that 2026 could end up being a better year overall for the nation’s lieutenant governors. Our historical analysis reveals that success rates for lieutenant governors running for other offices have not been constant from cycle to cycle. In fact, win rates rose almost constantly through the 2010s, such that in 2020 and 2022, over half of lieutenant governors running for another office won their races.

But that success rate plummeted in 2024, when only one of six lieutenant governors won their race (now-Gov. Mike Kehoe of Missouri). All that proves is that success comes and goes from cycle to cycle, sometimes in dramatic fashion. And that could mean that 2026 turns out to be a lucky year for the 16 lieutenant governors running for statewide or federal office, including Flanagan.

So far, though, Flanagan seems to be the exception rather than the rule. Up until this point in the cycle, lieutenant governors have struggled to break through in various races across the country. In most polls of Illinois’ Democratic primary for their open Senate seat, Lt. Gov. Julianna Stratton is trailing Rep. Raja Krishnamoorthi. In New York, Lt. Gov. Antonio Delgado just dropped his bid for governor amid faltering support. Low polling numbers in Michigan and California have pushed lieutenant governors to switch from running for governor to less crowded races for other statewide office. And just last fall, then-Lt. Gov. Winsome Earle-Sears lost her bid for governor in a landslide to Rep. Abigail Spanberger.

Finally, in Wisconsin and Georgia, lieutenant governors could end up having exactly a 50% success rate this year. That’s because each contest features a matchup between two lieutenant governors. In Wisconsin, Lt. Gov. Sara Rodriguez and former Lt. Gov. Mandela Barnes are facing off in the Democratic primary for governor. And in Georgia, Republican Lt. Gov Burt Jones and Democratic former Lt. Gov. Geoff Duncan are vying to win their respective party primaries for governor as well. That means at least two of the four will end up losing, depending on who makes it through their primary.

So when the dust settles on the 2026 midterms, it’s possible Flanagan ends up being one of the few rarified lieutenant governors to defy history, particularly when it comes to the Senate. We’re still months away from the August primary, more than enough time for Craig’s strong fundraising to turn the tide. But for now, Flanagan is defying the odds and crossing her fingers that 2026 turns out just a little bit better than usual for the nation’s second-in-commands.

Subscribe now

Share

Footnotes

In the State of the Union address on Tuesday, President Donald Trump made exactly one reference to polls. When discussing the SAVE America Act currently before Congress, Trump said the following:

I’m asking you to approve the SAVE America Act, to stop illegal aliens and others who are unpermitted persons from voting in our sacred American elections, the cheating is rampant in our elections. It’s rampant. It’s very simple. All voters must show voter ID. All voters must show proof of citizenship in order to vote. And no more crooked mail-in ballots except for illness, disability, military or travel. None. And this should be an easy one and, by the way, is polling at 89%, including Democrats, 89%.

So, is it polling at 89 percent? This week on The Trendline, we check in on how Americans are thinking about voting, democracy, and what parts of the proposed legislation they do and don’t support.

Voters are split on how our democratic experiment is going. In a Bowling Green State University/YouGov poll (Feb. 13 - 16, 2026), voters were asked “thinking about the future, how confident are you in American democracy?” Fifty-one percent said they were “very” or “somewhat confident,” while 50 percent said they were “not very confident” or “not confident at all” (rounding causes the numbers to exceed 100 percent). However, on this point, Democrats were more likely to express concern than other political groups: 36 percent of Democrats said they were very or somewhat confident, compared to 40 percent of independents and 71 percent of Republicans.

When it comes to the immediate future, though, voters seem to be feeling a bit better. In an Echelon Insights survey (Feb. 19 - 23, 2026), 63 percent of voters said they were very or somewhat confident that “the 2026 midterm election will be free and fair,” a position shared by majorities across the political spectrum: 69 percent of Republicans, 62 percent of independents, and 56 percent of Democrats agreed. However, in the same poll, partisans don’t really trust each other when it comes to elections. 66 percent of Republican voters said they thought that “Democrats are attempting to steal the 2026 midterm election,” while 77 percent of Democratic voters said the same of Republicans.

And while Trump cites concern about noncitizens casting votes in federal elections, that concern isn’t universal. A study from the Center for Transparent and Trusted Elections (Dec. 19, 2025 - Jan. 12, 2026) found that while trust that votes will be counted accurately has fallen, particularly among Democrats, the sources of that mistrust vary widely. Among Republicans, 51 percent said they had some or a lot of distrust about “people who are not U.S. citizens” casting ballots in elections, but only 11 percent of Democrats did. For Democrats, the greatest source of mistrust wasn’t about ineligible voters casting ballots or improper vote counting, but gerrymandering. Fifty-one percent of Democrats said that they had some or a lot of mistrust due to district lines not being “drawn in a way that fairly reflects what voters want.”

Now, on to proposed reforms. While the president mentioned a number of different reforms in his address, the SAVE America Act does not include all of those items. The key provision of the SAVE Act is to require states to “obtain documentary proof of U.S. citizenship and identity – in person – when registering an individual to vote in a Federal election.” In a Yahoo/YouGov poll (Feb. 9 - 12, 2026), Americans did support the requirement, but not as strongly as the president suggested. Sixty-two percent of respondents said they favor “requiring proof of citizenship – usually in the form of a passport or birth certificate – in order to register to vote,” while 23 percent said they would oppose the requirement. In the Echelon Insights poll, 67 percent of likely voters said that “requiring proof of citizenship to vote” would make the 2026 midterm election more fair, while 24 percent thought it would make it less fair.

And while Trump claimed bipartisan support for the proposal, according to the Yahoo/YouGov poll, that’s just not true. Democrats opposed the policy by 6 percentage points, 45-39, while Republicans supported it by 85 percentage points, 89-4. Independents also expressed support, but far less strongly; they favored the measure by 37 percentage points, 60-23.

Trump also mentioned other voting-related policies, such as limiting vote-by-mail. On this point, Americans don’t seem to agree with the president. In the Yahoo/YouGov poll, 46 percent of respondents opposed making it harder to vote by mail, while 38 percent were in favor. And even fewer supported making it harder to vote early in person: just 21 percent favor the policy, while 57 percent were opposed. In the Echelon Insights survey, 69 percent of likely voters said that “promoting early voting and secure mail-in ballots” would make the midterms more fair, while 20 percent said it would make them less fair – the opposite of what Trump is suggesting.

The most popular voting-related policy Trump mentioned was requiring voters to present identification in order to vote. In the Echelon Insights poll, this policy was the most favorable of all the policies they tested, with 76 percent of likely voters saying requiring ID to vote would make the 2026 midterms more fair, and just 17 percent saying it would make them less fair.

Finally, while it didn’t make it into Trump’s speech on Tuesday, there’s been ongoing chatter, including from the president, about whether the federal government should take over election administration, either nationally or in certain states. Voters oppose this policy strongly: in an ABC News/Washington Post/Ipsos poll (Feb. 12 - 17, 2026), registered voters opposed the policy more than 2-to-1, with 56 percent opposed and 25 percent in favor. Among both Democrats and independents, support was in the single digits, at 4 and 6 percent, respectively. Among Republicans, however, 46 percent supported the idea, with 24 percent opposed. This may be related to question wording, though, as the poll explicitly mentioned that Trump proposed the plan. With Trump’s name included in the question wording, Republicans may be more likely to indicate support than they would without that prompt.

Share

Subscribe to FiftyPlusOne to unlock the full polling experience on our website and get premium analysis that goes beyond the toplines. FPO subscribers get new analysis weekly and help keep the infrastructure for our website running.

Subscribe now

Other polling nuggets

• According to a new poll conducted by CNN/SSRS after Trump’s State of the Union address this week, just over half (54%) of people who watched the president speak said he was focusing on the right issues. That would be good news for the president, who, polls have shown, voters think is not prioritizing the issues that matter most to them, were it not for the fact that CNN also found this group of speech-watchers leaned politically to the right by 13 percentage points. (The sample size of Democrats in the survey was so small that the pollster was unable to report crosstab findings among them.) Among that already-friendly crowd, nearly half (45%) said Trump focused too little on the economy and cost of living, and just 31% expressed a lot of confidence in him to make living costs more affordable.
  
  CNN’s survey also found that speech-watchers were about evenly split on whether the president’s handling of tariffs was appropriate or an overstep of his powers. State of the Union addresses almost always get watched disproportionately by Americans who already approve of the president — in CNN’s polling dating back to the Clinton era, audience reactions have always been positive — but even grading on that curve, Trump’s 64% positive rating was lower than for the address he gave Congress last year, or Biden’s in his last year in office.

• A new survey from the Pew Research Center (Jan. 20 - 26, 2026) finds a 50-point partisan gap in who thinks their side is winning in politics — one of the widest Pew has recorded in a decade of asking the question. Six in 10 Republicans and Republican-leaning independents now say their side is winning more often than losing, while just 10% of Democrats say their side is winning — the lowest mark for either party since Pew first asked in 2016. Notably, even during Joe Biden’s presidency, Democrats were more likely to say their side usually loses in politics than to say they were winning.

• A new ABC News/Washington Post/Ipsos poll (Feb. 12 - 17, 2026) finds that while about nine in 10 voters who strongly disapprove of Trump say they’d vote for the Democratic candidate in the midterms, among those who only somewhat disapprove of Trump, 46% say they’d actually vote for the Republican candidate, compared to just 32% who’d go Democratic. This is a notable finding, as it suggests 2026 and 2028 could come down to whether Democrats can convert soft Trump disapprovers into actual Democratic voters. Disapproval of the president is a necessary, not sufficient, condition for the other party to win. However, just 10 percent of voters in the survey said they disapprove somewhat of Trump’s handling of his job (48 percent disapprove strongly). That’s a small sample size, so there’s a fairly large margin of error on this result.

• How do Americans feel about war with Iran? The latest Economist/YouGov poll (Feb. 20 - 23, 2026) finds that about half (49%) of Americans oppose the U.S. using military force to attack Iran, while just 27% support it. A majority of Republicans (58%) back such an attack, but majorities of Democrats (76%) and independents (54%) are against it, and even the 58% support among the GOP is lower than at the start of past U.S. military interventions. And yet, 58% of Americans think the U.S. military is likely to take action against Iran in the next month. These numbers suggest any military action would carry significant political risk for the administration heading into the midterms.

• Finally, from Echelon Insights’ February poll: voters apparently think vaping is bad but marijuana is good. The survey found that voters hold strongly negative views of e-cigarettes and vaping while being broadly supportive of marijuana legalization. Support for marijuana legalization now routinely clears 60% in national polls (Pew recently pegged it at 87% when you include medical-only supporters), while public attitudes toward vaping have soured amid youth health concerns.

Polling averages update

Texas Republican U.S. Senate primary

New polls released this week have pushed Texas Attorney General Ken Paxton up to a 5.3-point lead over incumbent U.S. Senator John Cornyn in the GOP primary for the Lone Star state’s Senate seat. Wesley Hunt trails both candidates, at 19% of the vote. Polls can be wildly off in primary contests, so it’s not impossible that Paxton could win 50% of the vote on Mar. 3 and avoid a primary — but it remains unlikely.

Texas Democratic U.S. Senate primary

On the Democratic side of the race, Texas state House Rep. James Talarico has recently pulled into a small lead against U.S. Congresswoman Jasmine Crockett, with 46.6% of the vote in our polling average vs Crockett’s 43.1%. That lead is within the margin of error and either candidate could win. We calculated that historical statewide primary polls picked the wrong winner in competitive races about 20% of the time.

2026 U.S. House generic ballot

Democrats have held their lead on the 2026 U.S. House generic ballot steady over the last week, at about 47% of the vote versus 42% for the Republicans.

Trump’s job approval

Donald Trump’s job disapproval hit a new high of 58.1% this week, while 38.9% disapprove of him. As of 12:00 PM on Feb. 27.

The poll, conducted by a firm called “Political Intelligence,” is not the first that the firm has done for Herrera. The first survey I noticed was conducted in December, and reported on February 13th by the Daily Caller. The second survey, which gained more attention, was reported by the New York Post on the 23rd. From there, it spread: it was retweeted by a reporter for the Texas Tribune, and further spread by news outlets like Axios and Politico.

Now, that’s all normal for normal poll coverage. But here’s the problem. I’ve spent two weeks trying to find out who conducts polls for “Political Intelligence.” I still have no idea, and the Herrera campaign hasn’t responded to my emails. I can’t email the company because I can’t find any information about them at all.

This is an issue for us here at 50+1 because we adhere to a strict set of methodological standards before aggregating polls from new polling firms. Part of that process is verifying the data came from a real pollster, was generated in a statistically sound way, and the pollster meets some minimal standards for disclosure like telling us who paid for the poll and the basic methodology by which it was conducted.

But in this case, we have been unable to find any of that information for “Political Intelligence”; the name of the firm is manifestly un-googleable (which isn’t actually all that uncommon in the shadier side of the polling universe). We don’t have any leads on who might be behind these surveys.

We’re all about transparency here at 50+1, and I felt this was a good opportunity to peel back the curtain on some of our research methodology. So without further ado, come on a journey with me as we try (and fail) to hunt down that which wishes to remain hidden.

The Hunt

Upon the release of the first poll on the 13th, I began a series of steps I would usually take when vetting a new pollster.

First, I reached out to my network of fellow poll hunters to see if any of them had any information. No dice.

Next, I tried googling, of course. But, dear reader, googling the phrase “political intelligence” yields all kinds of results, very few of which are useful. Go ahead, give it a try. After refining my search in various ways, I finally discovered a campaign firm called “Political Intelligence,” and thought I hit paydirt. But after reviewing their website, one thing seemed odd: this is a Canadian firm. Why would they be polling a Republican congressional primary in Texas?

However, this jogged something in my memory. Back in 2024, when I was working at FiveThirtyEight, I traced exactly these steps. I have seen Political Intelligence before, though the exact context is lost to wherever the archives of old FiveThirtyEight Slack messages are kept. I even recall reaching out to the Canadian firm to ask them if they conducted the poll, which they denied. I was unable to verify details about the company then, just as now. (For added security, I confirmed these details with a former coworker who also remembered that saga.)

Finally, in a last-ditch effort to get any information, I went through the Herrera campaign’s financial filings at the FEC website, to see if I could find any payments to Political Intelligence. Payments could include more information about the firm, like a physical address. But alas, it appears that many of the campaign expenses are going through a consulting firm, which is likely the entity hiring the pollster, rather than the funds being spent by the campaign account directly. This ensures that expenses remain opaque; the consulting firm is not required to submit filings to the FEC.

Look what you made us do

And that’s where I landed on the 13th. I decided at that time to just put it away as a mystery I may never solve, because the details I had from the Daily Caller article at the time were a bit thin, the poll was already two months old, and I wasn’t very optimistic about hearing back from the campaign: If a pollster is this hidden, it’s unlikely I’ll get any additional information.

Then came the 23rd, when the second survey surfaced in the New York Post. This one had a little more detail: attached to the report was a topline document. The new information here: a logo, which I promptly put through a reverse image search. I was hoping to turn up a website, but no such luck. All I revealed was a very short tweet referring to another survey conducted by the firm, with even less information than the other reports. But as to anything else, like a website or names of principal investigators, I seemed to be running out of moves.

Now, here is the dilemma. It is stated very clearly in the FiftyPlusOne methodology that per our ethical standards, “pollsters will disclose the names of their principals publicly.” That is to say, we are not in the habit of publishing surveys from completely anonymous organizations. There are rare exceptions to this rule (pollsters may have good reasons to ask me to keep their names off the record, and I may honor those from time to time), but on the whole, it’s important to us that the firms that we aggregate are on the up-and-up, and if we don’t know who is running them, we can’t vouch for that. However, given that this company appears to have been hired, routinely, by multiple campaigns, I do believe that this is a real polling firm conducting real polls that actually do ask questions to respondents and report their answers.

I wrestled with the question of whether or not to aggregate the polls. Other polling aggregators, who either know something I don’t know or perhaps have looser standards, have done so. (To my knowledge, we are the only aggregator that has a public set of standards for which polls we include, so I don’t know what methodology others are following.) Given the wider media distribution of the second survey, users of our site might be surprised or confused not to see it appear, and I do believe the surveys were conducted according to standard polling practices. On the other hand, it is explicitly prohibited by our stated methodology to aggregate data if we do not know who is behind it.

Ultimately, unless I hear back from the campaign, I have decided not to aggregate the polls. I feel badly about it, but I also feel that if standards exist, they should be followed. I trust our readers enough to understand why I made that decision, and hopefully this little missive helps you to trust us, too.

And finally, poll lovers, if you know anything about Political Intelligence, get in touch!


Correction: An earlier version of this article misspelled the name of Brandon Herrera as Brendan Herrera. The article has been updated with the correct spelling.

Subscribe now

The first competitive primaries of the 2026 election cycle happen in under a week, on March 3rd. In Texas, Democrats and Republicans will determine which candidates they want to face off against each other in November’s election to the U.S. Senate — one of the contests that could help determine control of the chamber at this year’s midterms.

For a smart primer on the candidates and where they stand on the issues, we’d point you to the Texas Tribune’s voter guides for the Democratic primary and the Republican primary. But 50+1 is here for the polls. And, if you’re a consumer of polls (like we are!), then there’s very little hard data on what is going to happen.

Where the averages stand

On the Republican side, it’s a three-way race. Attorney General Ken Paxton leads with 32.1% in our polling average1, followed by incumbent Senator John Cornyn at 29.2% and U.S. Rep. Wesley Hunt at 19.8%. That’s a Paxton +3 margin — a reversal from earlier in the cycle, when Paxton led comfortably before collapsing in the polls around September, around the time that Hunt entered the race.

With no candidate likely to crack 50%, this race is almost certainly headed to a May 26 runoff between the top two finishers. One hypothetical poll of a runoff between Paxton and Cornyn shows Paxton leading 51% to 40%. Over the last year, our average shows that Paxton’s margin declined after Hunt entered the race, so it’s likely the U.S. House representative is pulling from Paxton’s “lane.”

On the Democratic side, it’s a tight two-candidate contest between U.S. Rep. Jasmine Crockett and state Rep. James Talarico. In our average, Crockett leads in the polls by 42.7% to 39.7% — a margin of just 3 points, and well within the margin of sampling error for the polls provided.

Both races are close enough that the leader could easily lose, and in the Republican race, could even miss the runoff. And that brings us to the big caveat.

Primary polls are historically fairly unreliable

Frankly, you should rarely ever put too much stake into polls of primary elections. To put the averages above in the context of historical uncertainty in statewide primary polling, we looked at polls of Senate and Governor primaries held in 2014, 2016, and 20242 and calculated how accurate the polls were in each race. We computed a weighted polling average for 34 races (we only included contests where there were at least two surveys released in the last three weeks of the race), and the picture is, in a word, sobering.

At the individual candidate level, the historical error of statewide primary polling averages is 10 percentage points. That means a candidate polling at 31% could easily finish anywhere from the low 20s to the low 40s.

On the margin between the top two candidates — which is what we really care about in plurality races, like the Democratic primary in Texas — the numbers are even worse. The expected error on the margin between two primary candidates in a race for governor or U.S. Senate is a whopping 13.0 points. And in our historical data, polls picked the wrong winner for competitive primary races (those where the margin between candidates in polls was less than 10 points) in 2 out of every 10 elections. So both Crockett’s 3-point edge and Paxton’s 3-point advantage are well within the historical predictive margin of error for Senate primary polls.

But the average also obscures some truly enormous misses. In the crowded 2024 Nevada Republican Senate primary, polls had Sam Brown up 18 points on average over the last 3 weeks of the race; he won by 45. In the 2014 Iowa Republican Senate primary, Joni Ernst’s polling lead over her closest competitor was 16 points; she won by 38. And the 2024 Ohio Republican Senate primary polls showed a near-tie between Bernie Moreno and Matt Dolan; Moreno won by 18.

The 2024 cycle was also the least accurate of the three we studied, with an expected error of 16.1 percentage points on the margin of victory between the candidates polling in the top two. The 2014 cycle was the most accurate, at 10.8 points. Neither of these numbers should make you confident in a 3-point lead.

There are very few polls in either of these averages. The Democratic primary has only been polled a few times since late December, when former Congressman Collin Allred dropped out of the race, giving us barely two months of sparse data. And one out of the four polls of the Democratic race conducted in 2026 — accounting for 25% of the weight in the polling average — is a survey from the Crockett campaign. We have found that internal campaign polling tends to overshoot support for the campaign releasing the survey.

Share

Partisan polling on the Republican side

Speaking of pollster partisanship, an even larger share of the polling of the Republican primary has been conducted for organizations backing one of the candidates. This is fairly common in primaries, but the degree here is fairly stark. Of the 10 surveys conducted in 2026, only three or four came from neutral sources (one of them, a Cygnal poll from early January, was conducted for an unspecified candidate in a different race, and we don’t know if that candidate has endorsed in the Senate race). Four more were conducted for pro-Cornyn organizations, and the other two for pro-Paxton organizations. Looking at the entire length of the campaign, of the 40 surveys that asked about all three major candidates, just 12 came from neutral sources.

And, perhaps unsurprisingly, the sponsor of the polls makes a difference in the results. In polls conducted on behalf of organizations supporting Paxton, Hunt has never polled higher than 18 points, and Cornyn’s support has diminished to the high 20s over time. In polls conducted for pro-Cornyn organizations, Cornyn’s support has never dipped below 28 points, and in the most recent poll, he remains in the mid-30s. And in pro-Hunt surveys, Hunt’s support grew to the mid 20s by the end of 2025. In the few neutral surveys we have, there’s a bit of a question mark: two show the race as nearly a 3-way tie, while the other two show Hunt in the teens.

We can read some tea leaves into these polls, though. While pro-Hunt polls showed him in the mid 20s in December, no pro-Hunt group has published a survey conducted in 2026, which may indicate that his support has started to peter out: no news is bad news, in a way. And in the most recent pro-Cornyn polling, Cornyn is tied with Paxton, which may indicate that the best survey result the Cornyn backers are able to get is not really all that good for the incumbent. But ultimately, polls conducted for organizations with a dog in the fight are known to be less reliable than neutral polls, so we may want to take the entire polling average with a grain of salt.

Crosstab uncertainty on the Democratic side

Given the temptation for people to use primary polling data to argue about which candidate might be more electable in a general election, another concerning source of uncertainty in this race is in the polling crosstabs. A survey from Emerson College conducted for Nexstar Media, for example, has James Talarico leading in the Democratic primary by 30 points among political independents, 56% to Crockett’s 27%. Meanwhile, a survey from the University of Houston found Crockett leading independents by 18 points. In Texas, voters can choose which party primary to participate in, so appealing to independent voters is key to winning.

That’s a 48-point swing between two polls on the same crosstab — and it illustrates why you should be extremely cautious about drawing electability inferences from primary polling subgroups. Part of the explanation here is subgroup estimates are based on very small sample sizes, which can be noisy. The University of Houston poll above has 550 likely democratic primary voters overall, for example, which means it probably contains fewer than 100 self-identified independents. The margin of error on vote share among a group of just 50 people is 14 percentage points — and that’s per candidate, meaning the margin of error on the margin between Crockett and Talarico is close to double that. And that’s before you even get into the fact that the two pollsters may have defined “independent” differently or made different assumptions about who’s actually going to cross over into the Democratic primary.

There’s also a deeper methodological problem lurking inside the crosstabs: in an open-primary state like Texas, the composition of each party’s primary electorate is itself uncertain. If a surge of independent — or even Republican — voters shows up to vote in the Democratic Party’s primary (something that early vote data may already be hinting at) then the crosstabs from pre-election polls, which were calibrated to a different assumed electorate, could be especially misleading. There isn’t just uncertainty about how independents will vote, there’s uncertainty about what types of independents are voting in which party’s primary, and whether that’s the same group that will turn out in November.

This kind of thing matters, because Democratic primary voters are often considering general election electability as a factor in their decision. But arguments that Talarico or Crockett is clearly the stronger general election candidate based on primary crosstabs just can’t be backed up by the data.

Share

Polling primaries is just really hard

Part of covering public opinion polls is knowing when they are useful and when they are not. While polls give us good predictions of horse-race contests (especially presidential elections) on average, they are less predictive in primary elections.

This is in part because electoral dynamics can change quickly in primaries (candidates can drop out and endorse others), but also because primary polls are simply harder to conduct. Since the response rate to the average poll today is around 1%, pollsters rely on a statistical technique called weighting to make sure their poll is representative of the electorate by various demographic factors that are highly correlated with election outcomes, such as race, education, and party identification. But there are fewer highly predictive variables available to weight on in primary elections, where candidate support is more correlated with non-demographic factors like ideology, beliefs about issue prioritization and electability, etc.

Primary polls are also highly susceptible to decisions pollsters make about how to model who is likely to vote in each primary. For example, some pollsters base their turnout models on which voters turned out for previous primaries, while others ask voters directly if they plan to vote, but fail to sort out voting-aged adults who say they will vote, but are in fact very unlikely to. In races where vote intention is highly correlated with past participation — such as, for example, the 2025 New York City mayoral primary — these assumptions can also drive results (and errors). We’ve already seen a surge in turnout in the early vote in Texas, particularly on the Democratic side, which could be a warning sign for polls conducted with traditional, voter-file based turnout models.

In addition, primary polls tend to have significantly more undecided respondents than general election polls. In Texas Republican primary polls conducted in 2026, an average of 21 percent of respondents did not indicate which candidate they would vote for. These respondents may not be paying close attention to the race, but if they made it through the likely voter screen, there’s some indication, at least to the pollster, that they will vote. In some cases, respondents explicitly tell the pollster they plan to vote. In the Emerson College/Nexstar Media survey, for example, all the respondents told the pollsters they plan to vote next week, but only 71 percent expressed a preference for a candidate. How the remaining 29 percent of voters behave could mean the difference between who goes to a runoff, or if there is a runoff at all.

This all adds up to a lot of uncertainty about the polls ahead of next week’s elections. While your polling gurus at 50+1 are usually gung-ho about making inferences about politics with data, in this case, the empirical advice is actually to expect a lot of predictive error next week. Primary polls give us only a very rough sketch of these races. Paxton and Crockett technically “lead” in polls, but leads of 3 to 4 points are essentially coin flips given the historical accuracy of primary polls. The average 13 point error we see on polling margins means either race could produce a result that looks nothing like what the polls suggest. Don’t be surprised by a surprise.

Like this kind of analysis? 50+1 paid subscribers get premium deep dives like this one, 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

Footnotes

The explosion of investment into artificial intelligence has driven a surge in construction of new data centers all over America. This construction boom has contributed to some hesitance on the part of local governments and communities, and may have a political impact. In key states for the midterms, like Ohio, Texas, Pennsylvania, Wisconsin, and Michigan, political fights over data center development have erupted, often with scrambled party lines. In many states, leaders are openly welcoming new construction, while local communities are more skeptical. In Michigan, for example, Governor Gretchen Whitmer has been outwardly supportive of data center construction, even as more than a dozen municipalities and counties have passed moratoria on the projects.

As a result of recent artificial intelligence-related news, we’ve seen a new rash of polls looking into how Americans feel about the technology, and how they’re thinking about data centers in their communities. As it turns out, most Americans don’t really have a deep familiarity with AI, which could contribute to their data center hesitancy. For example, in a YouGov/Economist survey conducted this week, just 23 percent of Americans said they regularly use AI. Another 40 percent said that they have used it before, but don’t regularly use it. Nearly a third, 31 percent, said that they have never used it themselves (19 percent said they had seen it used, and 12 percent said they had never seen anyone else use it). And when it comes to trust in the technology, Americans appear to have a dimmer view. Just 7 percent said they have a great deal of trust in AI, 28 percent said they have a fair amount of trust in AI, but 58 percent said they had either not much trust or no trust in the technology.

Notably, these positions cut across party lines. Twenty-six percent of both Democrats and Republicans said they regularly use AI, and 30 percent of Democrats and 34 percent of Republicans said they had never used it. As to trust, 8 percent of Democrats said they had a great deal of trust in AI, a position shared by 9 percent of Republicans.

However, voters do see some value in AI. In an early February survey published by Echelon Insights this week, respondents were asked whether AI would have a positive impact on various aspects of American life. When it comes to areas like business, work, and shopping, Americans expect AI to have a more positive than negative impact. But respondents were much more negative about AI’s impact on education and cost of living. And overall, people are uneasy about the technology: a survey from Verasight, conducted in Dec 2025, found 56% of adults said they were anxious about AI tools.

And when it comes to jobs, people are especially pessimistic. In a survey from Cygnal conducted in early February, 9 percent of likely voters said that AI would create more jobs than it eliminates, compared to 62 percent who said it would eliminate more jobs than it creates (another 19 percent said that the number of jobs created and eliminated by AI would be roughly equal). Concern over job loss has led to a dramatic increase in the number of Americans who think the government should step in. In an Ipsos survey conducted in late January, 72 percent agree that “the government should take action to prevent the potential loss of jobs due to AI,” while 18 percent disagree. The last time Ipsos asked this question in April 2023, Americans were split, with 43 percent supporting government action and 42 percent against.

Now, the issue of data centers is a much more local one than that of artificial intelligence. In some communities where data center construction has been challenged or even halted, activists cite concerns about energy and water usage, tax revenue, and long-term viability of economic benefits, as well as general AI skepticism. And this local concern shows up in support for data center construction in polling, more generally: the further away a data center sounds, the more likely respondents are to support it. For example, consider three question wordings from three different pollsters. One asks about whether respondents would support a data center “in their area,” one asks if they would support one “in their community,” and the third asks if they would support one “within 10 miles of your home.” As the data centers get closer, support gets lower.

In addition, data center opposition based on energy costs does seem to be resonating with the American people. In a late January survey from Public First/Politico, among respondents who said they would generally support a plan to build a data center in their local area, support drops when costs increase. Asked if they would support or oppose a plan to build a data center in their local area, 37 percent of respondents indicated support, while 28 percent indicated opposition. However, among those that indicated support, that support drops significantly if the data center increases costs to local households. At a monthly costs to households of just $5, nearly a third of those who previously indicated they would support data center construction in their local area change their minds. At a cost of $50 per month, only a third are still on board.

Overall, people think artificial intelligence is moving too quickly. In a Fox News poll conducted in late January, 59 percent of voters said that AI was moving too quickly, while 6 percent said it was moving too slowly and 33 percent said it was about the right pace. Perhaps because of the speed of change, people slightly favor government regulation: in the Ipsos poll, 50 percent of respondents said that government should have a major role in the oversight of AI, while 41 percent said a minor role and just 9 percent said no role at all.

But even so, voters may not trust the government when it comes to technology regulation. In the Fox News poll, just 8 percent said they have a great deal of confidence in the federal government’s ability to properly regulate artificial intelligence technology, while 28 percent had some confidence, and a majority, 63 percent, had not much or no confidence at all. And to make matters worse, confidence in the federal government’s ability to regulate the tech has been decreasing over time. In April 2023, 39 percent of voters told Fox News that they had a great deal or some confidence, which has fallen to 36 percent in the latest survey. At the same time, the percentage who say they have not much or no confidence at all has risen from 59 percent to 63 percent.

Subscribe to FiftyPlusOne to unlock the full polling experience on our website and get premium analysis that goes beyond the toplines. FPO subscribers get new analysis weekly and help keep the infrastructure for our website running.

Subscribe now

Other polling nuggets

• Do people want government insurance for prescription drugs? YouGov did a survey experiment this week (Feb. 9 - 11, 2026) that tested support for the policy using two different question wordings. Half of respondents received a question asking if they approved of the government launching “a new website called TrumpRx.gov that offers discounts on some prescription drugs for people who pay without using insurance,” whereas the other half were asked about “a new website that offers discounts on some prescription drugs for people who pay without using insurance.” YouGov found that 41% of adults approved of the policy when they were given the TrumpRx condition, while support rose to 57% of adults on the generic policy description.

• Do you answer phone calls from strangers? YouGov also asked, in a poll conducted on Feb. 16, 2026, whether Americans answer phone calls from numbers they don’t recognize. Just 5 percent of adults said they always do, whereas 42% of the public responded that they never answer calls from unknown numbers. This is bad news for pollsters, who, just 30 years ago, could reliably get 30-40% of people to complete an interview when they called.

• A survey from RMG Research, conducted Feb. 17 - 18, 2026, found that a majority of U.S. adults favor banning “Wall Street Investors” from buying single-family homes, a policy some in Washington believe would help reduce housing prices across America. (The actual research on the subject finds less than 0.5% of single-family houses in the U.S. are owned by Wall Street, suggesting more limited effects of an investment-firm ban on house prices.)

• A new poll from Navigator Research (Jan. 29 - Feb. 1, 2026) probes the many reasons Americans hate congress. According to the survey, eight out of every 10 adults believe elected officials’ focus on enriching themselves and their donors stops them from getting things done, while 76% say members of Congress “[work] for big corporations instead of the people.”

• Speaking of corporations, SpaceX CEO Elon Musk recently announced the company would shift its strategic focus to establishing a colony on the moon over the next few decades. YouGov finds that 7% of Americans say they “would definitely” live on a lunar colony, while 48 percent “definitely would not.” Maybe response rates will be higher up there.

Polling average updates

Texas Senate primary polls

This week, FiftyPlusOne released new polling averages of the primary elections for Texas’s U.S. Senate election this fall. On the Republican side, Texas Attorney General (31.2%) currently leads Incumbent Senator John Cornyn (27.5%) in a 3-person race with U.S. Rep. Wesley Hunt trailing in third (20.1%).

On the Democratic side of the lone-star primary, U.S. Rep Jasmine Crockett currently leads state house Rep. James Talarico 42.7% to 39.7%. That three-point lead is inside the margin of error of our average.

Trump approval

According to the FiftyPlusOne.news average, 38.9% of U.S. adults approve of the way Donald Trump is handling his job as president today, while 57.2% disapprove. This time last month, his numbers were 39.6% and 56.8%.

Generic ballot

Democrats currently lead in our average of U.S. House generic ballot polls by 4.7 points, 46.7% for Democrats to 42.0% for Republicans. This time last month, Democrats were up 4.4 points, 46.0% to 41.6%.

Share

Subscribe now

Our polling aggregates have been cited in outlets including The New York Times, Axios, Politifact, Nate Silver’s Silver Bulletin, Washington Monthly, and MS NOW, and by public figures including Paul Krugman, Heather Cox Richardson, Jamelle Bouie, Philip Bump, and others. We offer poll lovers and casual political observers alike what we are certain is the fastest, most comprehensive aggregation of U.S. political polling in the world, as well as the only public portal for exploring raw data alongside polling averages that correct for known biases from different polling firms and modes. Our horse-race averages were more accurate than alternative methods at forecasting the results of the 2025 elections for governor in New Jersey and Virginia, and we’ll bring the same dedication to accuracy and rigor to the midterms and beyond.

Today, FPO is excited to announce the next steps in our journey to safeguard transparent public polling aggregation for the future: A more robust site with new features and a presence on Substack for editorial analysis of the latest polling data. Starting today, users will be able to log into our website to see complete tables of polling data (not just the most recent questions) on a wide range of topics, including Donald Trump’s approval rating, the generic congressional ballot, Senate primary elections, and more. We also have a new home page where users can see all of our active polling averages and recent articles.

And here on Substack, we will start publishing a weekly roundup of the latest polling data, plus weekly articles from our staff and contributors. We welcome pitches from academics, journalists, or anyone with an interest in public opinion research; reach out to editor@fiftyplusone.news if you’ve got a story idea you’d like us to consider.

Subscribe now

Today, we have also added polling averages for the Democratic and Republican primaries for the 2026 U.S. Senate election in Texas.

Should you want to see all the latest polls, your account for the FiftyPlusOne.news website will be tied to your account here on Substack. Simply subscribe to a monthly or annual subscription here, then go over to the FPO homepage and click “log in” in the top right corner. You will receive an email with a link to log into the website. It’s that easy.

A subscription to FiftyPlusOne.news is $8/month, or $80 per year. Paying subscribers will get access to premium analyses here on Substack, but to be totally up-front about expectations, the real benefit of the subscription is the data. Of course, subscribers also help ensure that transparent data aggregation will be sustainable for years to come. (For those that want access to raw data, API subscriptions are also available; reach out to data@fiftyplusone.news to learn more.)

In the coming months, we plan to expand the averages we offer to additional elections for the U.S. House, Senate and governors’ mansions, as well as polls about issues that matter to the public. In the summer, we hope to publish statistical forecasting models for the U.S. House and Senate, building on models that in 2024 significantly outperformed betting markets and other forecasters. All the data we publish will be available to premium subscribers of this Substack, with most of it also accessible programmatically to API subscribers.

We are still early on our journey to create a new, independent home for collecting and averaging public polling data for the masses. We are grateful to everyone who has followed our work from our initial alpha launch last fall, and to the outlets and writers who have helped put our data in front of a wider audience. If you believe that independent, transparent polling aggregation is worth preserving — or just want access to the highest-quality, fastest polling averages available — we hope you’ll consider subscribing and spreading the word.

Thanks for taking this next step with us!

— The 50+1 team

Subscribe now

Share

Go to the new 50+1 home page