Despite Wilkerson advantage, Chang-Diaz still ahead in poll

UPDATED: State Senate candidate Sonia Chang-Diaz sent waves through the local political community Monday when she released an internal poll that showed her holding a commanding 17.5 percent lead over state Sen. Dianne Wilkerson.

Sonia Chang-Diaz (D)

In an interview with PolitickerMA.com Tuesday, Chang-Diaz's pollster, Dan Cohen at Connection Strategies, said that not only was the firm careful about who it interviewed for the survey, but it used a turnout model that would, in theory, favor Wilkerson (D-Boston).

"We used a turnout model where we expect the largest percentage of turnout coming from those neighborhoods were Wilkerson did best," Cohen said.

In developing the model for the poll, Cohen had to determine how to weigh respondents from neighborhoods where Wilkerson did well and respondents from areas Chang-Diaz did well, based on the results of the 2006 primary. Wilkerson edged Chang-Diaz in that race.

When it came time to pick which model to use in determining the results of the poll, Cohen used a model that relied most significantly on neighborhoods where voters turned out to vote for Wilkerson in 2006 and other previous elections.

"We assumed strong turnout in Wilkerson's base within the parameters of previous elections," he said.

Cohen, however, also noted that if he had used another model, one that skewed towards Chang-Diaz, for example, it wouldn't have made much of a difference. Maybe one or two points, he said.

Cohen said he decided to use this model because the campaign "wasn't sure we'd be releasing the results of the poll. We needed an accurate picture so we knew where to focus our effort in the final stretch."

The Wilkerson campaign declined to comment for this article.

David Paleologos, a pollster at Suffolk University, didn't question the poll's results. Paleologos said the most important factor in these polls is determining the likelihood that the respondents are going to vote.

"You need to establish voter intensity," he said. "And you can define likely voters in two ways."

First, Paleologos said, the pollster can look at the voter history and determine if the voter cast a ballot in recent similar elections. And second, the pollster can ask the respondent the likelihood that he or she will vote.

Cohen said he used both standards. His company called voters that had voted in the last recent low turnout Democratic primaries. Then the questioner asked the respondent at the beginning of the interview how likely they were to vote in September.

The survey "only continued with people that said they will definitely or probably vote," Cohen said.

Ultimately, Cohen said he believes the poll shows that the convention wisdom in the race - that Wilkerson is leading - may be unfounded. But that doesn't surprise Cohen.

"I've been working on elections professionally since 2000," he said. "And especially with Democratic primaries, conventional wisdom is a coin toss if it's accurate or not. I've never bought into the conventional wisdom."