Mark Blumenthal | March 22, 2007
Thanks to MyDD, we know that the Service Employees International Union (SEIU) released poll results yesterday based on samples of likely primary or caucus voters in Iowa, New Hampshire, Nevada and South Carolina (summary, state by state results, results by party). The MyDD post was, unfortunately, a bit confused about the meaning of some of the state-by-state results. Since MyDD (and political junkies everywhere) are appropriately focused on polling in the early primary states, let's try to help clear up the confusion.
The survey was conducted by the Democratic firm Lake Research Partners** and had a somewhat unique design: They interviewed 1,607 likely primary and caucus goers in each state, with roughly 400 per state, 200 likely voters from each party in each state. Also, the sampling procedure was different for most other public polls:
Telephone numbers for the sample were drawn randomly from a statewide voter file, with primary/caucus voters selected on their vote history.
Pollsters disagree on the merits of sampling from voter files , but for better or worse, samples of past primary voters make for a very different population of respondents than other public polls on the 2008 presidential primaries. As noted in my two recent posts on the subject, national polls on the party primaries tend to include all adults (or sometimes registered voters) that identify or lean to a given party. In this case, the pollsters interviewed only those with some past history of participation in primaries or caucuses, and then (see the questionnaire) screened all but those certain or probable to attend either the Democratic or Republican caucus or primary in their state.
Now all of this would make for an interesting test of methodologies, except that this survey did not include traditional vote preference questions. And that's where things apparently got confusing. MyDD diarist robiliberal initially reported what looked like horserace results for the Democrats in each state. On promoting the post to the front MyDD page, Chris Bowers updated with this comment:
OK, I just realized that this poll only included people who view health care as one of their top issues. That explains why Obama does so bad everywhere--young voters are one of his bases, and young people just don't usually include the cost or availability of health care as one of their top issues (I actually learned that at a union organizer training session with SEIU). This isn't about all Democrats, and so is thus less useful. But still interesting, and the kind of poll we could do of netroots types--Chris
Not exactly. The samples of Democrats included all those who qualified as likely primary or caucus attendees as per my description above. What was different is the text of the question:
Who do you think would be the BEST president on the issue of health care?
So all Democratic primary/caucus goers got the question, but it was a question about health care. Needless to say, that's not exactly, "for whom would you vote if the election were held today." Not surprisingly, Hillary Clinton - who led the task force on Health Care reform during her husband's presidency - has a special advantage on that issue that results in a bigger "lead" over Obama than on pure vote preference measures.
**Interest disclosed: I worked for Celinda Lake for two years in the early 1990s.