Tuesday, January 13, 2009

Alaska Senate: Polar Opposite Polling

By Nathan L. Gonzales

A newly released Dittman Research (R) poll showed Sen. Lisa Murkowski (R-Alaska) leading Gov. Sarah Palin (R) by a wide margin in a hypothetical 2010 GOP Senate primary matchup. The results were also the complete opposite of another public poll taken during the same time period.

Murkowski led 57 percent to 33 percent in the Dittman survey, taken Dec. 5-20, and paid for by the Alaska Standard, a Republican Web site. A Dec. 15-17 survey by the nonpartisan Research 2000 for the liberal, Democratic Web site DailyKos.com showed Palin leading a primary 55 percent to 31 percent.

Even though Palin defeated Lisa Murkowski’s father, the incumbent governor, in the 2006 gubernatorial primary, there is no indication that a 2010 matchup will materialize.

“In my opinion, Alaska is hard to poll accurately,” Dittman said, “Many outside research firms have problems here — Rasmussen is the exception, they have a very good record — but Kos is one of the worst. In the recent general election here, the final Kos survey was probably the most inaccurate poll in the nation.”

“One thing we learned in 2008, it’s that Alaska is clearly a difficult place to poll,” said Markos Moulitsas of DailyKos.com about the Senate and House races. “The state wasn’t kind to any pollster last year.” But Moulitsas did point out the accuracy of Research 2000 at the presidential level in Alaska.

Maryland-based Research 2000 showed Sen. John McCain (R) with a 58 percent to 39 percent lead in an Oct. 28-30 survey. Anchorage-based Dittman Research showed McCain with a 56 percent to 37 percent advantage in an Oct. 24-29 poll. And Rasmussen had McCain up 16 points on Oct. 28. The GOP nominee won the state 59 percent to 38 percent.

Pollsters inside and outside the Last Frontier took their lumps last year in Alaska’s downballot races.

In the House race, no public polls, dating back as far as fall 2007, showed Rep. Don Young (R) ahead of Democratic challenger Ethan Berkowitz, and most had him trailing by at least 5 points. Young, who is under investigation, was re-elected 50 percent to 45 percent.

In the Senate race, most of the polls gave Anchorage Mayor Mark Begich (D) a slight to narrow advantage over then-Sen. Ted Stevens (R) throughout the race, but there were only a couple of public polls after the Senator’s conviction on seven counts of corruption.

An Oct. 28-30 Research 2000/DailyKos.com (D) poll gave Begich a 22-point lead, Rasmussen gave Begich an 8-point advantage in an Oct. 28 survey, and Alaska-based Hays Research showed the mayor with a 7-point edge. Begich won the race 48 percent to 47 percent.

According to Dittman, his firm has correctly predicted every primary and general election winner for Senate and governor of Alaska for 34 years, even though its margin in last year’s Senate race was also wider than the final result. A Dittman poll on the House race was not listed on Pollster.com.

The strange thing about the 2010 Senate polls is that the difference is not really one of margin. The results are totally opposite, and the explanation is not clear.

Dittman expressed concern about the Research 2000 methodology in local media stories, but declined to give specifics.

His poll surveyed 505 adults, with an over-sample of 430 Republicans for a sample for the primary question, from Dec. 5-20 (much longer than the traditional research period). The survey was paid for by the Alaska Standard, a conservative, Republican Web site that lists Dittman and Lisa Murkowski as contributors. The site’s publisher, conservative talk show host Dan Fagan, is a well-known Palin critic.

Research 2000 surveyed 600 likely voters, including 400 likely GOP primary voters for the Senate question, from Dec. 15-17. The poll was paid for by Kos Media LLC, but Moulitsas does not have a hand in the way the survey is conducted and posts all crosstabs and results on his Web site. Research 2000 President Del Ali noted in an interview that screening and predicting likely primary voters almost two years before an election is difficult and a potentially fruitless exercise.

Even though the ballot question results were different, the surveys agreed that Palin and Murkowski are popular.

In the Research 2000 poll, Murkowski had a 51 percent favorable/43 percent unfavorable among all voters (including 72 percent favorable/22 percent unfavorable among Republicans), compared to 60 percent favorable/38 percent unfavorable for Palin among all voters and an overwhelming 88 percent favorable/10 percent unfavorable among Republicans.

Dittman Research tested job approval instead of person favorability. Seventy-six percent of all respondents gave Murkowski a “very good” or “quite good” job rating compared to 17 percent “not too good” or “pretty bad.” Palin held a lower 65 percent “very good” or “quite good” job rating and 34 percent “not too good” or “pretty bad.” Job numbers among just Republicans were not available.

For now, there are two very different polls for a race that will likely never exist. And in general, Dittman appears unwilling to specifically criticize the methodology because that would then give away the secret recipe to polling the state. So who is right? We’ll probably never know.

“I have no interest in a pissing match,” Moulitsas remarked. “Aggregating multiple polls gave us great predictive power in 2008 in pretty much every race except Alaska Senate and Alaska at-large. I want more polling. I see no point in trashing anyone doing reputable work.”

This story first appeared on RollCall.com on January 9, 2009. 2009 © Roll Call Inc. All rights reserved. Reprinted with permission.