National Journal.com

nationaljournal.com > Hotline On Call

Daily Kos Dismisses Pollster

The liberal website DailyKos.com has chosen to terminate its relationship with MD-based pollster Research 2000, according to DailyKos founder Markos Moulitsas.

"I have decided to part ways with our current pollster, and will be looking for a new polling partner to finish out this election cycle," Moulitsas wrote in a statement posted on his site.

Research 2000 Pres. Del Ali confirmed the split in an interview with Hotline OnCall.

Moulitsas' decision comes in the wake of last night's AR SEN Dem primary runoff, in which Sen. Blanche Lincoln held off LG Bill Halter, Daily Kos' preferred candidate. The website promoted Halter as a replacement on the Dem ticket months before he announced his candidacy, and the site's readers were encouraged to contribute to his insurgent campaign.

Research 2000 conducted two polls for the site between the primary and the runoff. While Halter posted narrow leads in both polls, Ali noted to Hotline OnCall that those leads were within the polls' margin of error.

"It's not 'Dewey Defeats Truman'," Ali said, "but that's the way it goes."

AR SEN wasn't Research 2000's only high-profile "miss" of the primary season. In a poll conducted May 17-19, Research 2000 showed Rep. Artur Davis (D-AL) leading his state's GOV primary over Ag. Commis. Ron Sparks (D), 41-33%. But Sparks trounced Davis, 62-38%, in the June 2 primary.

Ali told Hotline OnCall that criticism of the AL survey was unfair because it was conducted two weeks before the election. Indeed, part of the problem may have been the dearth of public polling in the race; the Research 2000 poll was the only live-caller poll received by The Hotline in the year preceding the election.

AL isn't the only state in which Research 2000 has fared poorly in pre-primary polling. In their final CA SEN GOP poll, which was conducted back in mid-May, ex-Rep. Tom Campbell led ex-Hewlett Packard CEO Carly Fiorina by 15 points. Fiorina downed Campbell by 34 points last night, less than three weeks after the poll.

In his final KY SEN primary poll, conducted a week before the election, Research 2000 had opthamologist Rand Paul (R) leading Sec/State Trey Grayson by 10 points, but Paul won by 23.

Moulitsas said his decision was also based, in part, on new "pollster rankings" published earlier this week by Nate Silver, the baseball statistician whose politics website, fivethirtyeight.com, was recently licensed by the New York Times. Silver's site ranked Research 2000 among the worst three independent, live-caller polling firms that had conducted more that 10 public polls in WH, SEN, GOV or House races since '98.

"Research 2000's past results aren't actually as bad as people will try and make them out to be," Moulitsas noted in his statement.

"In fact, if there's one thing that's striking about the chart, is how closely clumped together those pollsters actually are. The difference in accuracy between the best and worst pollsters ... isn't very big."

Ali claims that Silver hasn't disclosed all the data on which the rankings are based.

"I can't really speak to what he does because, frankly, nobody knows," Ali told Hotline OnCall.

In a methodology statement posted on his blog, Silver said the database on which the rankings are based contains 4,670 distinct polls. Silver did not immediately return an e-mail seeking comment.

In general, Research 2000 polls have tended to favor Dems this cycle in general election matchups, due in part to the pollster's undersampling of older voters compared to other polls and past exit poll data. When those sampling and weighting differences were reported in this space in early Apr., Moulitsas defended Research 2000.

"Del certainly isn't one to base existing demographic guesses on historical ones, and in 2008, he was pretty spot on," Moulitsas said at the time.

That faith, however, seems to have been eroded over the past few months.

Moulitsas wrote on his site today, "I believe in accountability, in accuracy, and in making sure we provide the absolute best information not just to this wonderful community, but also to the outside world."

The polling banner atop the DailyKos homepage was quickly removed, but, as of press time, the results of past Research 2000 surveys are still available on the web server.

Research 2000 is among the most prolific public, live-caller state pollsters thus far this cycle, conducting roughly two or three state polls per week for Daily Kos in recent months, in addition to polls for other media outlets or outside groups. Ali told Pollster.com's Mark Blumenthal, in a column for NationalJournal.com earlier this year, that roughly 80% of the polling by his firm that was made publicly available was done for Daily Kos or other progressive groups.

19 Comments

Such a cool blog. I liked this post especially the best.

So Markos wants a poll with more propaganda. How funny. ;0)

Use your crystal goblets.Do not save your best perfume,and use it every time you feel you want it

Pollster needs to dump Rasmussen.

Actually KansasGirl, the piece specifically quoted Moulitsas about how he's committed to accuracy, accountability and provide "the absolute best information". Research 2000 was not providing that for him. I mean, correct if I'm wrong....

First, the article says: "Research 2000 polls have tended to favor Dems this cycle in general election matchups, due in part to the pollster's undersampling of older voters compared to other polls and past exit poll data"

Gee, anyone guess in off-year elections what demographic has one of the highest turnout and typically over performs?

Second, though I didn't access specific data, but for simplicity sake let's say the county is 35D/30R/35I. My guess is they tend to oversample democrats and assume independents haven't fallen off the Obama wagon yet and will continue their voting pattern from 2008.

Finally, since they are undersampling older voters, they must be oversampling younger voters, those 1st time voters who voted for Barack and assume they'll vote.

Anyone guess which demographic typically has lower turnout and under performs in off-year elections?

This is what my college business stats teacher referred to as SISO: $hit In, $hit Out.

But I give Moulitas credit. If the polling isn't correct, or at least close, then one cannot determine where to focus resources for maximum effect to change an election. Just ask the SEIU and Bill Halter out in Arkansas.

I downloaded The Orange Box torrent from www.games-iso.com a couple of weeks ago and, even though its a"dusty" game, its the best NINTENDO DS game ever made in my opinion =)

Ask and you shall receive. Nate has posted a new article with every Research2000 poll that he had in his database. Ali should have kept his mouth shut.

http://www.fivethirtyeight.com/2010/06/on-transparency-hypocrisy-and-research.html

good share, great article, very usefull for us¡­thanks!

I love reading and I am always searching for informative information like this Great content

We love our President, and do not believe the polls. The real polls are grass roots. Grass roots polls will re-elect President Obama in 2012! Believe it! We are what is called the silent majority.

Intriguing site. My co-workers and I had been just discussing this the other evening. Also your page appears excellent on my aged sidekick. Now thats uncommon. Great perform.

Hi there could I reference some of the insight found in this blog if I reference you with a link back to your site?

with ‘we leave it to you to decide’.

generally, nicely, the truth hurts.

I come across your blog on technorati and check a couple of of your early posts. Keep up the great work. I just added up your RSS feed to my MSN News Reader. Looking forward to reading more from you down the road!

The site above suggests you can pick up a free 3DS on the net.. Has anyone tested out any of those free Nintendo 3DS offers????

I would like to say "wow" what a inspiring post. This is really great. Keep doing what you're doing!!

this site needs to put captcha on the comments - there's way too much spam here