Polling Firm Released Fake Data in Hong and B*** Races as a ‘Social Experiment’

An organization that released public election polling in recent weeks withdrew the results of two surveys, revealing that it had fabricated the polls as part of a “short-term social experiment” meant to expose how quickly false information could spread.

One of the surveys released by the group, Median Strategies, claimed that Francesca Hong was leading the Wisconsin Democratic primary for governor by more than 20 percentage points. (She went on to lose by less than one percentage point.) Another, shared by Mayor Karen B*** on X in a since-deleted post, showed Ms. B*** leading Nithya Raman by nearly 12 percentage points in the Los Angeles mayoral runoff.

On Monday, Median Strategies withdrew those results, saying they “should not be treated as genuine polling data.”

In a statement posted to the organization’s website, Median Strategies said the fabricated results were intended to examine “how purported polling information could enter and spread through the political information ecosystem without independent verification.”

Median Strategies did not respond to a request for further comment on the incident.

The false polls, particularly the one from the California race, spread widely after being picked up by The California Post and various social media polling aggregators. The results, however, were not published by multiple major polling aggregators, and were not included in The New York Times’s poll trackers because the group did not respond to questions about its methodology.

The experiment came as the number of public pollsters has skyrocketed, with around 30 new pollsters cropping up so far this year. The large number of surveys coming from new, previously untested pollsters has made it increasingly difficult to judge survey quality. But while these new pollsters may be unproven, the vast majority appear to be real.

In Wisconsin, the Median Strategies results were largely in line with the other public polls, all of which greatly underestimated Ms. Hong’s opponent, David Crowley, igniting a wave of criticism of the value of primary polling.

What Median Strategies hoped to accomplish by adding false polls into a saturated marketplace was unclear, even to experts in the field.

“It doesn’t make sense as a social experiment; what are you measuring?” said Mia Costa, a government professor at Dartmouth College. “How do you know who was exposed and who wasn’t? This is just wreaking havoc in an election with real stakes and no real interest in understanding the effects of what you are doing.”

The revelation of such an experiment, and how easily the false results spread, exposed a challenge in a polling world where incentives are high to produce polls that attract attention, and the cost of a low-quality poll is relatively modest.

“The bar to doing a poll has gotten so low that it’s easy to get through without adhering to typical industry standards,” said Josh Pasek, a political science professor at the University of Michigan. “As a result, these bad actors can quickly manipulate our understanding of a race.”

The fake polls, which were first reported on by The Los Angeles Times, also drew attention to the increasing reliance on low-quality polls and betting markets to inform the public’s understanding in races where little information is otherwise available. Additionally, it calls into question how information of varying quality might move the needle in betting markets like Kalshi and Polymarket.

“The ability to manipulate even a marginal amount of information can have a disproportionate impact on the markets,” Dr. Pasek said.

Ms. B***’s election odds showed slight improvement in the betting markets after the Median Strategies poll was reported, while no changes were seen in Wisconsin. In its statement, Median Strategies claimed that no one involved received any compensation or financial benefit from the project, and that none of those involved held or traded any position in the various prediction or betting mechanisms.

Without further explanation, the statement about the experiment ended with a message that stood in contrast to the confused reaction that followed from people throughout the polling world: “Thank you for your understanding.”