Scott Milne, the Republican gubernatorial candidate, has recently cited two polls as evidence that he is gaining on Governor Peter Shumlin, a Democrat. A New York Times/CBS/YouGov.com opt-in, online poll of 430 Vermonters conducted Aug. 18-Sept. 2 shows a 11 percentage point gap between Milne and Shumlin. In the survey, 41 percent of Vermonters supported Shumlin, and 30 percent backed Milne. Fifteen percent of the respondents said they were undecided. A previous robo call survey by Rasmussen Report, also conducted in August, gave Shumlin a 12 percent lead over Milne.
Eric Davis, a retired professor of political science from Middlebury College, questions the methodology of both polls.
The gold standard for polling, according to Davis, is based on three criteria: interviews with live operators, landline and cellphone calls, and weighting based only on demographic factors, such as age, gender or race.
Davis says the Castleton Polling Institute, Washington Post and Pew Research Center all use these standards.
Rasmussen falls short in all three areas, Davis said. The company uses robo calls instead of live operators, doesn’t include cellphone users in its surveys, and weights polls based on the partisan history of the state, he said. There is no way of knowing if people who respond to the Rasmussen calls are over the age of 18 or registered voters.
Because of the partisan weighting, Rasmussen tends to have a 3 percent to 5 percent Republican bias, Davis said. In 2012, the company predicted that Mitt Romney would win the presidency.
“I would not place much credence in polling methodology with the flaws Rasmussen has,” Davis said.
The YouGov.com poll is a self-selected online response from individuals and is by no means a random sample.