Party ID assumptions and polling

Obie1948

Community (This content is not subject to review by Daily Kos staff prior to publication.)

Sunday October 05, 2008 · 7:01 AM PDT

2008/10/05 · 07:01

The Star Tribune poll poll out of Minnesota has Franken way ahead of Coleman for US Senate and Obama way ahead of McCain. It appears they did not follow accepted polling methodology of assuming a predetermined party identification differential. Nate Silver at FiveThirtyEight.com discusses this poll and this issue, and he concludes that he would not trust the poll, and "The Minnesota senate race, most likely, is a toss-up."

Nate may well be right to question the methodology of the Trib poll, but what if ...

For comparison, we have the SUSA poll in Mn which shows McCain plus 1 (compared to the Trib poll of Obama plus 18!). The SUSA poll assumed a differential in party ID of plus 7 for the Democrats over the Republicans. Based on this assumption, their raw data of polling responses was adjusted to fit this preconceived model. This is apparently accepted polling methodology and national models typically assume a plus 6-9 point advantage for the Dems and raw data is adjusted to fit these assumptions.

The Trib poll made no such assumptions and simply accepted the respondent's answers without adjustment. The raw data indicated the respondents identified themselves as Democrats at a plus 16 rate over those who identified themselves as Republican.

As expected, the Republicans have criticized the poll for that reason as does Nate Silver. But the Trib's pollster argues that their methodology reflects reality.

Forty-two percent of respondents identified themselves as Democrats in the latest poll, up from 34 percent a month ago. In the same period, the percentage of Minnesotans calling themselves Republicans has dropped from 31 percent to 26 percent.

Larry Hugick, chairman of Princeton Survey Research Associates International, the Star Tribune's polling firm, has studied party identification shifts in presidential years and found that it moves in concert with the results of the candidate horse race. In a study that he and an associate conducted during the 2004 presidential race, they found that 18 percent of registered voters changed their self-described party affiliation between September and October 2004.

What if this pollster is right? That would suggest that the national shift to Obama is understated by adjusting the raw data to preconceived assumptions of party ID. Just saying ...

By the way, cell phones were not included in this poll.

This content was created by a Daily Kos Community member.

Make YOUR voice heard!

Comments are closed on this story.