I am still playing catch up on a bunch of interesting new
surveys that all seemed to be released during my vacation last week. Next up: The
survey
of likely Iowa
Democratic caucus goers sponsored by the anti-poverty ONE Campaign and
conducted by a partnership of two partisan campaign pollsters: Peter Hart &
Associates (D) and McLaughlin & Associates (R). After we posted
and linked to the results, one reader emailed with a few good questions:
Is a privately commissioned poll on
a particular subject, in this case poverty, considered the same as an independent primary poll?With the preceding questions all
leading to one most likely conclusion, given Edwards' poverty platform and his just having concluded a poverty tour, is this poll pushing?
First, the reader got the wrong impression about the order
of questions from the ONE Campaign release,
which first presented results from questions about poverty, health and the ONE
Campaign itself and then presented results from the presidential trial-heat. The
actual questionnaire was different: I checked Geoff Garin of Peter Hart &
Associates and he confirms that they asked the presidential trial-heat question
first, before any of the items about poverty or America's role in the world. They
also confirm that the introduction that respondents heard at the beginning of
the call was general and did not make any specific reference to poverty or the
ONE campaign. So there is no reason to assume that the nature of the questions
"pushed" respondents to any particular candidate.
Second, to answer a question asked by another reader,
neither Peter Hart nor any of his associates are currently working for any of the
Democratic candidates for President in 2008.
So do we consider "privately commissioned" polls as
equivalent to "independent" surveys sponsored by news media outlets? We do
include both in our averages, but make sure to note sponsorship and the
partisan affiliation of the pollster, if any, and urge readers to take
sponsorship into account. One of our main goals for Pollster is to provide a
complete aggregation of all publicly available trial heat results as well as
the tools and commentary to try to make your own decisions about the worthiness
of the various polls. We monitored
the impact of partisan polls on House district averages during the 2006 general
election, and hope to do more of the same going forward.
One question readers did not ask, however, involved the
sampling methodology used by the ONE Campaign poll. Their release described the
respondents as:
509 past Democratic caucus attendees or new registrants
who say they are likely to attend a Democratic presidential caucus.
Does that blurb imply the use of a list of past
caucus goers? Yes. Via email, Geoff Garin confirms that they drew their sample
from the list of registered voters maintained by the Iowa Democratic Party that
identifies past caucus participants. Virtually all of the survey's 509 respondents
had actually participated in either the Iowa Democratic caucuses in either 2004 or 200. The exceptions were 3% of the respondents who had registered since 2004 and indicated on screen questions that they planned
to participate in the 2008 caucuses.
The approach of the ONE campaign survey thus stands
in marked contrast to the recent ABC
News/Washington Post poll that used a random digit dial methodology
to first contact a sample of all adults in Iowa and then screen down to self-described
likely caucus goers. According to the Washington Post version of the questionnaire, 30% of the respondents indicated that they had not previously participated and that the 2008 caucuses would be their first.
I want to look more closely at the contrast (and the
strengths and weaknesses of each) in a subsequent post, but for now, one
conclusion seems obvious: These two surveys produced very different results
that probably result from the way each pollster sampled "likely caucus goers." John
Edwards receives 30% of the vote on the ONE survey and runs significantly ahead
of both Hillary Clinton (22%) and Barack Obama (18%). On the ABC/Post
survey, Obama (27%), Clinton (26%) and Edwards (26%) are tightly bunched,
although the results get closer to consistent when they look at a "more
restrictive likely voter definition" (showing Edwards 28%, Obama 27%,
Clinton 23%).
So which approach is best? Getting to a definitive answer is probably impossible, but my hunch is that the reality of "likely caucus goers" falls somewhere in between. More on that soon...