Vermont media is atwitter this afternoon with new polling numbers on the Democratic primary for attorney general.
So what do the numbers tell us? Perhaps nothing.
The poll, which surveyed a random sample of 477 registered voters earlier this week, will certainly grab some headlines. It shows incumbent Bill Sorrell wielding a commanding 44-to-24 edge over challenger TJ Donovan.
But in a late-August primary that by most accounts will be lucky to draw 12 percent of registered voters to polls, how reliable is a poll of 477 random registered voters?
“I have serious doubts about the validity of this poll,” says Eric Davis, professor emeritus of political science at Middlebury College.
Even Rich Clark, director of the Castleton Polling Institute, which conducted the survey, says he wouldn’t put a ton of stock in the results.
“If the election was held by calling up people and asking their preference, Sorrell kills right?” Clark says. “But that’s not what’s going to happen. We’re going to have a low turnout and whoever turns out voters wins. I doubt the Donovan campaign look at this and says, ‘it’s over.’”
Primary voters are a tiny subset of the electorate, one whose voting behaviors differ from larger universe of voters in general elections.
“But I don’t have good scalpel to cut out those who are least likely to vote,” Clark says. “Polling in a primary race, especially where you have low turnout, is extremely difficult for that reason.”
Davis says the he’s troubled by one of the poll’s stats in particular – the percentage of respondents who say they’re “very likely” to vote in the primary. According to the poll, 59 percent of registered voters say they’ll turn out next Tuesday.
But turnout in the five-way Democratic gubernatorial primary in 2010 was only 24 percent, and most political observers say the Democratic primary this year will be lucky to draw half that. Poll respondents are notorious for inflating their likelihood of voting. ButDavissays the 59 percent number stands out nonetheless.
“I have serious doubts about the validity of a survey which is based on a group of respondents of whom nearly 60 percent say they’re going to vote when the actual turnout is likely to be closer 10 percent,”Davissays.
Clarksays he thinks it’s a “closer race than the numbers show.”
“But if I was betting, I would bet on Sorrell,” he says. “We don’t turn away incumbents that often in primaries. And Sorrell does better with older voters, which are also more likely to turn out in primary.”
According to Clark’s poll, Sorrell also does better among voters in Chittenden County, where turnout will be highest next week. He’s also favored over Donovan among voters who voted in the 2010 primary.
“So that tells me he’s got some support among the habitual voters,”Clarksays.
The poll found that Donovan, 38, has the advantage among younger voters and women.
Davis, however, says he has a hard time putting any stock at all in the findings that deal with sub-categories of respondents. When the sample size falls below 200, Davis says, margins of error become widen enough to render the result meaningless.
As for the poll’s 44-24 findings in favor of Sorrell?
“I don’t believe that for one second,” Davis says. “My guess is it’s going to be very close next week.”
The poll found that 31 percent of would-be primary voters are undecided.
The poll also tapped the pulse of Vermonters on general election races. If the gubernatorial election were held today, Gov. Peter Shumlin would defeat his Republican challenger by a margin of 60-to-25. President Barack Obama, meanwhile, retains a 62-25 lead over Mitt Romney. Both results mirror the outcome of the CPI’s last Vermont-related poll, conducted in late February.