The Barrett Campaign's Uneven Polling Criticism
By Steven Shepard // May 30, 2012 | 5:24 p.m.
Milwaukee Mayor Tom Barrett's campaign is upset by the new Marquette Law School poll released on Wednesday that showed him seven points behind in his quest to unseat Wisconsin's Republican Governor, Scott Walker, in June 5's recall election. But the reasoning behind the campaign's complaint is inconsistent with its praise of a Democratic poll it says is more accurate.
Barrett's communications director, Phil Walzak, released a statement about two hours after the Marquette poll was made public, calling it "an outlier."
"The MU poll predicts the electorate to resemble 2010," Walzak wrote in the statement emailed to reporters, "which is reflected by its undersampling of younger voters."
Walzak's statement cites a number of Democratic campaign polls released over the past week-and-a-half showing a slightly closer race. "Most of them show the race either tied or separated by just a couple of points," said Walzak. "This includes a poll released today that shows the race in a dead heat, 49%-49%."
Walzak is referring to an automated poll conducted by the Democratic polling firm Lake Research Partners for the pro-labor group Greater Wisconsin Committee. That poll was first reported by the Washington Post's liberal blogger, Greg Sargent; automated polls do not meet Hotline On Call's standards for publication.
But if the Marquette poll is undersampling younger voters, the Lake robo-poll ignores them almost entirely. Respondents to the automated Lake poll were asked, using their touch-tone phone, to enter their age near the end of the survey. Just 2 percent of respondents to the Lake poll said they were between ages 18 and 29. Meanwhile, in the Marquette poll, 16 percent of registered voters -- and about 12 percent of likely voters -- fell within the 18-29 age range.
According to the 2010 exit poll, 15 percent of voters were between 18 and 29 years old; they broke for Barrett, 55 percent to 45 percent, despite his single-digit loss. In 2008, which the Barrett campaign thinks is a better analog for next week's recall vote, 22 percent of voters were under age 30, and they voted for President Obama over John McCain, 64 percent to 35 percent.
Why does the Lake robo-poll -- the poll to which the Barrett campaign is clinging and even using as a fundraising tool -- undersample young voters so egregiously? Most automated polls undersample younger Americans because automated polls cannot call respondents on cell phones, according to federal law. The latest data from the National Health Interview Survey, from the first half of last year, show that nearly six-in-ten Americans aged 18-29 lived in households without a landline telephone, almost twice the rate among all American adults. Celinda Lake's office said she was traveling out of the country Wednesday afternoon.
Marquette Law School visting professor Charles Franklin, the poll's director, told Hotline On Call that his poll, which did include respondents contacted via cell phone, is weighted for age, and that their target for registered voters was 16 percent.
Asked whether the fact that the percentage of young voters decreases to 12 percent after applying the likely-voter screen means that their poll is undersampling that group, Franklin said, "I think we're in the ballpark. We know that the youngest voting cohort is the lowest turnout."
The Barrett campaign also claimed that the Marquette Law School poll oversampled "some of the reddest counties in the state." According to Marquette, 13 percent of respondents were from the city of Milwaukee, while another 33 percent lived within the Milwaukee media market, though outside of Milwaukee proper. Seventeen percent were from the Madison media market, 18 percent from the Green Bay/Appleton market in the northern part of the state, and 19 percent were characterized as being from the rest of the state.
Franklin said respondents' location was recorded when they told interviewers the county in which they live. He did say that the 33 percent who lived in the Milwaukee media market but outside the city -- traditionally Republican areas -- was slightly more than the statewide figures for all registered voters. "It's not perfect," said Franklin, "but, goodness, I'm perfectly happy with that match."
"I don't think it's fair to say we're way off on the region," Franklin added.
The geographic breakdown of the Lake automated poll differed slightly: 15 percent Milwaukee, 24 percent "outer Milwaukee," 17 percent Madison, 27 percent "Northeast" Wisconsin and "Northwest" Wisconsin.
A Barrett campaign spokesperson did not immediately return a phone call seeking clarification on their statement.
Leave A Comment