The trouble with the Star Trib poll, cont’d

In 2002 I wrote a column exploring the trouble with the Minneapolis Star Tribune’s Minnesota Poll. It was a poll that seemed regularly to overstate Democratic strength. Rob Daves was the Star Tribune employee responsible for the Minnesota Poll results I was writing about. In 2002 I offered to bet Daves dinner for two at a restaurant of the winner’s choice that Norm Coleman would do at least five points better than Daves’s final pre-election Minnesota Poll would show. Daves wisely declined the bet; Coleman did 9 points better than Daves’s final pre-election poll results.
Daves no longer works for the Star Tribune. The Minnesota Poll is now conducted “under the direction” (whatever that means) of Princeton Survey Research Associates, the outfit that performs widely derided polls for Newsweek. Over the weekend the Star Tribune published a Minnesota Poll showing Democrat Mark Dayton leading Republican Tom Emmer 39-30 percent. Independence Party candidate Tom Horner was at 18 percent.
The partisan breakdown of the poll sample showed Republicans at 28 percent, lower even than the 2008 exit survey that showed Republicans in Minnesota at 36 percent. (Democrats in the poll published Sunday were at 35 percent.)
The Weekly Standard’s Jay Cost is writing this week on how to read polls; his column today is here. I asked Jay if he could give us any guidance on the current Minnesota Poll.
Jay queried whether the poll’s partisan breakdown applied to the likely voter portion of the sample as opposed to the larger sample including “adults” that the poll uses for other purposes. Jay pointed out that the poll had about 9 percent of respondents belonging to a “third party” or “other,” yet the poll’s cross-tabs lack such a category. He wondered whether that means they were included in the Independent category or were simply excluded from the cross-tabs. “If you you assume that their partisan split is of likely voters and that the third-party people are in the Independent category,” Jay wrote, “their clear over-sample of Democrats gives Dayton about a two-point bump.”
The invaluable RealClearPolitics has also posted recent poll results from Rasmussen Reports and SurveyUSA, both of which show the race to be neck and neck. The Minnesota Poll results are an outlier in this group. How does Princeton Survey Research Associates stack up against Rasmussen Reports and SurveyUSA? Princeton is perhaps best known for the polls it conducts for Newsweek, which are widely thought to be something of a joke.
UPDATE: A reader points out that Newsweek was judged to have come in dead last in accuracy in an assessment of 23 2008 presidential polling organizations published by Fordham University. Rasmussen and Pew tied for first.

Recommend this Power Line article to your Facebook friends.

Responses