Remember: our rules allow pointed disagreement, but require respectful discussion.
--
I don't pretend to be a polling expert or an expert in statistical
procedures, but I'll take a try at answering the question raised by
Tim Bonham: Is the Star Tribune's screening of likely voters valid?
For a more technical explanation of polling techniques, call Rob
Daves, who runs the poll here at 673-7278. He can talk for hours
about polling techniques, and he will if you let him. He also does
peer review for Public Opinion Quarterly, the scholarly magazine for
people who poll.
Tim asks if the Star Tribune adequately screens to determine which
voters are likely to vote, and whether self-reporting produces invalid
results. Here's the procedure:
First, the confirmed non-voters * people who say won't vote in the
upcoming election -- are excluded. Then because people tend not to
admit that they won't vote, a four-question screening is used to
detect who is more and less likely to vote. For example, people might
be asked if they are registered, if they voted last time, if they'll
vote in the next election, and so on. They are then ranked on the
basis of their answers. More weight is given to the poll responses of
people judged more likely to vote, and less is given to the people
less likely to vote. A critical component is that this weighting is
based on historical research. It's not practical, given
considerations of timeliness, to go down to City Hall in the midst of
a poll to see if 400 people actually are registered or did vote last
time. So this weighting is based on the past voting performance of
people who scored similarly on the screening questions. In other
words, after a group of people is screened and polled, whether they
actually voted is researched and sorted by their screening rank The
proportion of these people who actually voted determines the weighting
given to people who will be screened and polled the next time. There
are still flaws in this, such as in 1998 when a larger proportion of
the gubernatorial general election vote was accounted for by
less-likely voters or people who registered on election day, both
groups tending to favor Ventura. But it's based on sound polling
methodology.
Steve Brandt
Staff Writer
Star Tribune (Mpls-St. Paul)
425 Portland Ave.
Minneapolis, MN 55488
612-673-4438 (voice)
612-673-4359 (fax)
e-mail: [EMAIL PROTECTED]
_______________________________________
Minneapolis Issues Forum - Mn E-Democracy
Post messages to: [EMAIL PROTECTED]
Subscribe, Unsubscribe, Digest option, and more:
http://e-democracy.org/mpls
