Over the last six weeks, three polls were released to the public in the OR-1 race. We now have election results data to compare them to.
In mid-December, Public Policy Polling (PPP) conducted a poll on behalf of Daily Kos and SEIU. At the end of December, SurveyUSA conducted a poll on behalf of KATU. And on January 11-12, Republican pollster Bob Moore (pdf) conducted a poll for Rob Cornilles.
That's right, folks. Despite being the closest to the actual election by weeks, Bob Moore's poll was wildly off in comparison. (Note: This post was updated on 2/2 at 11 a.m. to reflect the latest election results.)
A few more points of distinction: Sure, Moore's sample size was smaller and margin of error bigger. But PPP and SUSA were also much more transparent. Both released the exact wording of every single question, along with detailed demographic crosstabs. And PPP, as with every poll sponsored by Daily Kos, released the raw answer-by-answer data. Bob Moore? He just released a memo that characterized Cornilles as "surging" and the race "going right down to the wire".
This isn't the first time, of course, that Moore's polls have been called into question.
A week before the election in 2008, Moore had Gordon Smith beating Jeff Merkley by 4%, when two other polls conducted that same day had Merkley up by 5% (Hibbits) and 7% (SUSA). Merkley, of course, won by 3%.
In 2010, Moore had Scott Bruun beating Kurt Schrader by 4%, less than a week before voters started casting votes in an election that Schrader won by 5%.
And most dramatically - in 2010, he did it in the race between Jim Huffman and Ron Wyden, claiming Huffman was leading 47 to 38%.
The O's Jeff Mapes called that poll "a fantasy", and national pundit Stu Rothenberg was apoplectic - noting that Moore refused to release any details and saying that the campaign "has something to hide." Wyden won handily 57 to 39.
So, it seems to me that serious journalists shouldn't bother listening to Bob Moore's nonsense. And if they do, they should insist on complete transparency - all the questions, crosstabs, and raw data. Otherwise, they might just be getting taken for a ride through fantasyland (and taking their readers along with 'em.)