Charles Djou is blaring the news. The main headline on his campaign website this week announces: “We have an early lead in our Congressional Re-election Race!”

The item links to a story published Monday on the Hotline On Call blog, part of the National Journal, a nonpartisan political publication. The post says a poll from Djou’s camp shows he “starts the general election in a rather strong position,” leading Democrat Colleen Hanabusa 50 to 42 percent. The poll results “would cast a bit of a cloud” over Dems’ predictions that they’ll take Hawaii’s First Congressional District back in November, Tim Sahd writes.

The story briefly mentions that the poll was conducted by the Tarrance Group. What’s left out, or at least glossed over, is any explanation of what that means. Djou paid the Tarrance Group to conduct the poll. It’s not independent. And his camp won’t say what the questions were. Well, the incumbent may be winning, but questions can skew answers, so any poll where you don’t know what was asked has to be suspect.

While most people understand that polls commissioned by a candidate should not be taken at face value, Djou’s team is touting the results for all to see, so it’s worth rehashing the potential problems.

The Virginia-based Tarrance Group describes itself as a “Republican strategic research and polling firm,” and says it has helped elect more than 80 Republican governors, U.S. Senators and members of Congress. Brian Tringali, the partner who sent the poll’s results to Djou, was once director of the Research Division at the National Republican Congressional Committee (NRCC), according to his bio.

A memo from Tringali and a Tarrance Group research analyst to the congressman titled “Djou Shows Majority Support As New Freshman” was provided to Civil Beat by the NRCC on the condition that it not be republished in its entirety but instead used as background for reporting — otherwise we’d make it viewable right here. The NRCC said the poll is internal to the campaign and used to make strategic decisions but was leaked to the media to tout its positive results. Presumably, no poll results would have been leaked had they not been positive.

They leaked the poll, a blogger wrote about it, then they linked to the article the blogger wrote, rather than the internal memo that inspired the coverage. They used a news organization to give their own poll more legitimacy than it otherwise would have had. That distinction wouldn’t have been obvious to an unsuspecting visitor to Djou’s website.

The National Journal’s story earned a link from the National Review Online, a conservative news analysis site, with a bad pun for a headline and a revealing question: “Does any Republican running for the House this year not have internal polls that look great? Anybody at all? No? Okay, glad that’s settled.” (emphasis in original) The National Journal post was also republished in its entirety by the conservative Hawaii Free Press, though no mainstream Hawaii media appears to have picked up the poll.

So what does the memo say, anyway? On July 26 and 27, Tarrance conducted a telephone survey of 400 registered First Congressional District voters likely to participate in November’s general election. A random sample of that size would yield a margin of error of plus or minus 4.9 percent in 95 of 100 cases. This means that even Djou’s alleged eight-point lead, with 8 percent of likely voters still undecided, is sizable but not necessarily secure.

In addition to not releasing the questions, the Djou camp wouldn’t talk about how the Tarrance Group chose which interviewees would best represent the electorate. The poll could have been carefully manipulated to elicit certain responses, but Sahd, the National Journal blogger who wrote about the poll for a nationwide audience, said he’s confident that the poll was done professionally and accurately.

“That’s because The Tarrance Group is one of a handful of respected Republican pollsters here in Washington that dozens of Republican campaigns across the country depend upon for solid polling analysis,” Sahd wrote in an e-mail to Civil Beat when asked if he trusted the poll’s validity. “A professional firm like Tarrance Group does not engage in ‘push polling,’ so I have no worries that this survey was done in that manner.”

Sahd uses what’s fed to him to craft an analysis of the race. But he also gives Djou the headline he’s looking for.

If the general election were held today, would Djou really cruise to a victory? We don’t know, but Djou doesn’t know either.

Sahd made a good point in his e-mail to me: “It’ll be interesting to see if state Senator Colleen Hanabusa releases her own survey to counter Djou’s. If she doesn’t, we may find out that
her numbers look similar to his.”

But that approach turns politics into dueling polls, polls that can’t be trusted.

About the Author