More fake polling organizations were uncovered this week. Wedgewood Polling announced they deliberately fabricated certain polls as “an experimental project to demonstrate the vulnerabilities, lack of oversight and overall low barrier to entry in the modern political polling industry.” Several examples of their polling in states like Michigan and Wisconsin showed progressive candidates either leading or in close races that turned out to be very different than the actual result. Another polling outlet, SoCal Strategies, has suddenly disappeared after questions about its legitimacy.
One of Wedgewood’s late June polls showed El-Sayed well ahead of Haley Stevens (46-39) with Mallory McMorrow at 9% in the Michigan Democratic primary; the final margin between El-Sayed and Stevens was only 1 point. This was not the only poll showing El-Sayed with a larger lead than he had in the primary, but this poll clearly fed the narrative that he was dominating the primary.
McMorrow blamed the Wedgewood poll as negatively impacting her campaign and “that killed us.” She said, “Donations dried up, Volunteers were devastated…Reporters hounded me with calls to drop out.” Her take was confirmed in a post by a Democratic operative who ran a pro-McMorrow SuperPAC: “This poll was used by several donors to pull back funding at a time we were being outspent by $80 million.”
Wedgewood was among the polls showing Wisconsin candidate Francesca Hong leading in the gubernatorial race; the final result was that Hong narrowly lost to David Crowley, the candidate that the Wedgewood poll had shown in single digits.
Unfortunately, this is not a surprising development. The use of survey data has been moving away from understanding to validation, As these fabricated examples have shown, there is clearly a market for what they were offering. We have written before about ways to spot a fake poll or a poll with significant bias. Here are three red flags to look out for:
1.) Demographic information is unavailable or limited to prevent verification. Full demographics should always be available especially in a campaign poll. There can also be problems with a pollster’s weighting decisions and if their sample accurately reflects the voting electorate. If demographics are not available, it is impossible to evaluate a pollster sample.
2.) All of the data points are favorable to a particular narrative or agenda, likely for the agenda of the poll sponsor. If every number is favorable to one side, it should be obvious that the poll is intended for validation, as the less favorable data points are being hidden or were not asked to generate the most favorable outcome.
3.) A data report sounds like a press release than an analysis. Consisting of only favorable data points, the report reads like a press release to promote a point of view, and doesn’t provide a complete picture. There is no testing of competitors’ arguments or inclusion or any data points that might run counter to the narrative being promoted. These polls are geared toward advocacy or fundraising, or in a campaign context, trying to show that a candidate can win.
Polling problems have been perpetuated by lack of accountability. Once an election is over, the political world moves on to the next election without reviewing what happened. Faulty polling is likely to become an even more prevalent trend in the 2028 cycle, as campaigns and special interests will be looking to shape political narratives. The more people can be educated about red flags to spot, the more accountability can be introduced into the public domain.
For more on the problems in polling, see our podcast episode Why are polls wrong?
