This week’s Pollwatcher features trends in drinking among less religious Americans, views on live-streamed executions and support for allowing wagers on elections. But first …
The election polling fire hose is on full blast three weeks before the midterms, with seemingly countless polls gauging support in congressional, gubernatorial and even state legislative races. Who is conducting these polls? How are they carried out? Are they trustworthy?
Midterm polling this year looks a bit like the Wild West, with a huge variety in sponsors, methods, track records and transparency.
At least 59 different pollsters have released more than 150 polls measuring national support for Congress or support in competitive Senate contests from September to early October, according to the New York Times.
Most public polls are conducted by research companies or similar organizations (69 percent), although a sizable share are conducted by universities, media, political campaigns or parties. Many polls involve more than one type of sponsor, such as Washington Post-Schar School state polls or Fox News polls conducted with a team of Democratic and Republican polling firms.
There’s also a mix of election polling experience, with 74 percent of 2026 polls conducted by pollsters that have conducted at least 10 election polls close to general elections in the past, according to the Silver Bulletin. The typical poll is conducted by a firm with 87 previous election polls; that figure is much lower for state-level U.S. Senate pollsters (26) than for national pollsters (61).
National and state polls use strikingly different methods to reach voters. Most national polls draw samples from opt-in online panels of survey takers, while another 22 percent use online panels recruited through random sampling (like Post-ABC-Ipsos polls) and 21 percent use text-to-web methods.
By contrast, most state-level U.S. Senate polls use text messages (72 percent) and live interviewers (59 percent) to contact potential respondents, with 43 percent using opt-in panels. These percentages add up to more than 100 because many use multiple methods. State polls also are nearly twice as likely to use random or “probability-based” sampling methods, such as drawing samples from a state voter registration database.
Transparency is spotty: 93 percent of pollsters disclose the survey’s mode of interview (e.g., online, live caller), and 78 percent report any specific weighting factors. But just over half, 55 percent, specify where the poll’s sample came from, such as a voter registration database or an online panel. State polls are particularly poor on this front, with 46 percent disclosing the source of their poll’s sample, compared with 71 percent of national polls.
Eighty-six percent of national and competitive Senate polls came from groups that named staff members in their poll releases, but 14 percent came from firms that did not. Several pollsters do not appear to name human staff members on their websites, including Quantus Insights, SoCal Strategies and Wedgewood Polls. This specific lack of transparency was one of the factors that helped the Los Angeles Times identify that polls from the firm Median Strategies were fake this summer.
Want this in your inbox?
Sign up for The Pollwatcher newsletter here.
These patterns are a mixed bag for quality
The diversity of poll sponsors and methodologies is generally a plus, putting different approaches to conducting polls to the test. If polls of many different methods show similar results, that should give more confidence in the findings. But the big variety in methods also reflects the decades-long breakdown of traditional polling methods, failures of polling in recent presidential elections and the vast differences in financial investment.
Live-caller phone polls — still the preferred method for many top public and campaign firms — are quite expensive, often costing tens of thousands of dollars. Opt-in panel and text-message polls are far cheaper, and some firms have succeeded in producing accurate election polls with these methods. But opt-in panels have struggled with accuracy in representing the broader population, while pure text-message surveys are getting their first big test this year.
The scattershot transparency of polls is a bigger concern, undermining the ability of survey research methods to improve. It’s difficult for pollsters to understand what methods work best if firms don’t disclose basics about how they are conducted.
Worth your attention
Secular Americans are drinking less: Polls have long shown that less religious Americans drink more, but a new report from Gallup suggests that trend is changing. Among those who seldom or never attend religious services, the percentage who said they drink alcohol fell from 74 percent in 2019-2020 surveys to 58 percent in 2025-2026. Drinking also dropped by nine points among those who attend services nearly weekly or monthly, while they remained lower among weekly worshipers (44 percent). As Gallup’s Frank Newport notes, Americans have grown much more likely to say alcohol consumption is bad for their health in recent years, a period which overlapped with a pandemic spike in consumption.
Eight in 10 oppose live-streaming execution of Fort Hood shooter: A Washington Post-ABC News-Ipsos poll finds a narrow 53 percent majority of Americans support executing Nidal Malik Hasan by firing squad, while 42 percent are opposed. But 8 in 10 Americans say the execution should not be live-streamed to the public. On Thursday, President Donald Trump canceled plansto live-stream the execution.
Most Americans oppose election prediction markets: A Reuters-Ipsos poll this month found 60 percent of Americans opposed the use of prediction markets to wager money on U.S. elections, while 13 percent supported this and 24 percent said they were not sure. The poll found more than 6 in 10 opposed wagers on U.S. military actions and government policy announcements or decisions. Opposition was lower for wagers on professional sports at 43 percent. A separate question found 75 percent of adults think wagers on the outcome of real-world events is a form of gambling, while 8 percent said it was not gambling and 14 percent were unsure.
What do you want to know?
Send your questions, comments and ideas to [email protected].
The post Welcome to the Wild West of election polling appeared first on Washington Post.
