Pollsters weather storm of fake data, secret sponsorships, with some afraid to give names
Shortly after a mysterious new firm published a poll about the mayor’s race in Los Angeles, a small account on X raised a red flag. The pollster’s website is “brand new,” and its social media account “just started posting yesterday.” The skeptical user, @TheWuhanClan, concluded: “This account, poll and domain looks like a hoax.”
Minutes later, that account posted screenshots from Grok, the platform’s native AI model, which came to a similar conclusion about the firm, Median Strategies. The firm’s lack of contact information and sudden appearance, “is highly atypical for a credible firm and matches patterns seen with short-lived or fabricated polling accounts.” That night, Washington Monthly Politics Editor Bill Scher, who had already been raising concerns about faulty polling, casually observed, “Another brand new pollster enters the arena.”
Download the Straight Arrow app today to get the stories that matter free from manipulation, bias or agenda.
Point phone camera here
It would be another seven days before Median Strategies on Aug. 17 would acknowledge its poll was bogus, claiming in a statement to the Los Angeles Times that the firm was created as part of “a short term social experiment,” to reveal “how purported polling information could enter and spread through the political information ecosystem without independent verification.”
On the same day Median Strategies released its poll, Aug. 10, it also created its eponymous website, according to data reviewed by Straight Arrow News, a detail that could have raised questions about the firm sooner had anyone looked.
Close observers of campaigns have long grumbled about the lack of high-quality local polls, which have gotten more expensive as fewer people answer telephone calls, particularly from unknown phone numbers. That has led some campaigns to release internal polls they commission. And the financial incentives of newly created prediction markets have only complicated the hunt for reliable public polling. The result has been a tug-of-war among number-crunching websites, between aggregating large numbers of polls and selectively highlighting ones with more credibility.
Median’s poll did not get much attention from reporters or major polling aggregating sites, but its admitted deception is “another hit to public trust in polling at a time when our industry is already fighting a credibility battle,” Mike Noble, a pollster in Arizona, said in a statement to Straight Arrow.
Days before Median admitted its deception, Scher, the editor, had already seen enough faulty numbers flying around. He published a column on Aug. 14 with the purposely provocative headline, “Should We Ever Trust Polls Again?”
In an interview Tuesday night, Scher told Straight Arrow that X was the “Wild West,” because of the amount of questionable polling data circulating there. “Election Twitter,” Scher lamented, “sends everything around.”
Median is not the only firm whose polls have raised concerns.
The popular poll-aggregating site FiftyPlusOne announced Wednesday it was taking action against two firms for violating ethical and research standards. Not only are the two firms banned from the site, but so are “future firms founded or managed by” the principals at those firms, it said.
The first firm, The Public Sentiment Institute, received a $1,575 payment from Republican gubernatorial candidate James Fishback of Florida on June 9, according to public records cited by FiftyPlusOne. The following day, the firm published a poll about the race without disclosing its affiliation with the candidate, the aggregator said. On June 30, TPSI published another poll about the race with a similar omission, it said.
More alarming, the site said, was the firms’ decision to change results it got from certain respondents whose “profile” better aligned with Fishbacks. FiftyPlusOne said this is “tantamount to falsifying data.”
The second firm, Patriot Polling, received a $1,500 payment on Jan. 21, also from Fishback. It released a poll on Feb. 2 featuring Fishback and only one other candidate in the race, saying in a memo, “we have decided to conduct a head-to-head poll.” FiftyPlusOne described this as a misleading omission.
Fishback and Patriot Polling did not immediately respond to Straight Arrow’s requests for comment. In an unsigned statement to Straight Arrow, TPSI acknowledged its failure to disclose the payment but said “we strongly reject” the claim of data manipulation. “No respondent was removed, no response was altered, and no data was fabricated,” the statement said. An “error” occurred when it “applied an unvalidated adjustment layer” to the data “and failed to run a diagnostic” that would have caught the error, it said.
While it acknowledges the disclosure and methodological errors, TPSI said neither, “amounts to deliberate data falsification, and the underlying record does not support that conclusion.”
Another anonymous firm says it wants its privacy.
Wedgewood Polls’ website has no employee names, phone number or even an indication of where it is located. To get to Wedgewood’s website, visitors have to type “sites.google.com.”
Its sole social media account, on X, was created in June. Since then, it has released several polls collected on sites like FiftyPlusOne, the right-leaning MacIver Institute, and the polling aggregating pages of The New York Times, which notes, “This pollster does not meet our criteria for selected pollsters.”
One longtime Milwaukee journalist last month accused Wedgewood of being “a shady outfit” masquerading as progressive. But Adam Carlson, a Democratic pollster at a different firm, told Straight Arrow on Thursday he communicated with someone at Wedgewood after he raised a question about them on X. Carlson said the person told him the two-person firm includes one progressive and one conservative, and that, “They passed the smell test on terminology and things like that.”
In response to inquiries from Straight Arrow, they said they are a two-person shop, as Carlson described. Wedgwood declined to reveal their identity or appear on camera, but did respond to several questions, and acknowledged that their anonymity raises suspicions.
“We understand that this decision can naturally raise skepticism,” the firm said in one statement, “particularly given recent concerns surrounding organizations such as Median Strategies, Patriot Polling, and The Public Sentiment Institute.”
Public information about that eponymous site shows it was registered on June 26, one day after it released a poll about the Senate race in Louisiana and while it was conducting polling for the Senate race in Michigan.
The firm said it is “not currently incorporated,” but hopes to be soon. It says it generates revenue “through private polling work for individual clients.” One of the two people at Wedgwood “has previously experienced doxxing and online harassment,” leading to “the decision to keep the individual identities of those involved with Wedgewood private.”
The firm declined to participate in a video call.
Scher said people often forget polls are meant to be a “snapshot in time,” and not “what an outcome is supposed to be.” But that misperception, plus the increasing difficulty of getting a representative sample of people to respond to polls, is making the work all the more challenging.
“You can do all the right things and not get it right at the end,” Scher said. After two decades of looking at polls of various quality, he said he understands if pollsters ask themselves, “What’s the point … Why do it?”