Wedgewood Polling admitted this week that nearly all its published surveys were fabricated, after its figures appeared in election averages and coverage from major outlets. The organization described the deception as an experiment testing how readily a fictitious pollster could acquire credibility.

Its surveys largely favored Democrats, including Abdul El-Sayed in Michigan’s Senate contest against Republican Mike Rogers. Wedgewood said it conducted one genuine Nebraska poll as a baseline, but often supplied numbers unsupported by actual interviews. The New York Times has since removed 15 Wedgewood surveys, maintaining that it had questioned the organization about personnel and methodology before accepting its answers. Decision Desk HQ removed 17 surveys and reran its forecasts, reporting no change in projected congressional seat totals or chamber-control probabilities.

VoteHub said a conversation with someone identifying himself as a college student with polling experience had persuaded it to include Wedgewood’s work. Another aggregator, FiftyPlusOne, said its questions prompted the disclosure. Wedgewood denied intending to influence an election, although its admission nevertheless established that invented results had circulated through institutions readers relied upon to distinguish research from unsupported political assertions.

A polling average often looks more authoritative because it combines results rather than trusting one survey. Yet that protection crumbles once the underlying inputs are invented. While aggregation can smooth genuine sampling errors, it cannot authenticate interviews that never occurred.

Publication then supplies a second layer of credibility, as readers encountering a number in a respected outlet are far less likely to investigate the firm behind it. Once other organizations repeat the result, repetition can resemble corroboration even though every citation rests on the same fabrication.

The Democratic tilt warrants scrutiny without establishing that editors knowingly favored false results, but the documented failure is more concrete, as plausible answers about methodology were accepted as sufficient assurance that research existed. A questionnaire became a substitute for verification.

Decision Desk’s unchanged forecasts limit claims about the numerical damage. They do not erase the editorial failure. Coverage can influence perceptions of momentum without changing a model’s final allocation of seats.

Calling the exercise an experiment offers no exemption from responsibility. Wedgewood manufactured evidence, and publishers distributed it. Both helped turn unsupported political claims into something presented to the public as measured opinion.

Fabricated polling has been a recurring problem this election year. While a glass-half-full perspective might appreciate that they admitted the true nature of their ‘experiment’, it also shows that manufactured results can gain authority simply by reaching a publisher willing to put them into circulation. A skeptic might question why the fake polls largely favored Democrats, and why they came clean less than two weeks before the midterm elections.

These numbers, whether real or fake, can distort judgments about which races deserve donations, volunteers, and attention. A supposed Democratic surge can pressure Republicans to abandon positions or divert resources before anyone establishes that the voters behind it exist.