[extropy-chat] [psl at acm.org: ACTION ALERT: Blackwell purgedOhio Voter Rolls Oct 1st.- Vote Early!]
transhumanist at goldenfuture.net
Sat Oct 21 15:08:12 UTC 2006
>We go on and on about cheating at the election booth, but what about
>cheating in the opinion polls before the election? Those opinion polls may
>effect the outcome of an election, yet as far as I know there is no law
>against falsely reporting the outcome of an opinion poll, or designing an
>opinion poll to come out a certain way, or to select respondents with at
>least 10 sets of earrings. (Conservatives generally avoid punching holes in
>their bodies. Any person with more than two earrings or with earrings in
>any body part other than the earlobes are not conservative voters.
>Pollsters can easily select or deselect these if they wish.)
As a rule, the more transparant the pollster is willing to be concerning
their methodology, the more you should be able to trust their results.
If they are unwilling to publish the actual questions, or provide the
complete script of a poll on request, that should raise red flags.
Membership in AAPOR (the American Association of Public Opinion
Research) is also a good indicator of a reputable polling outfit. AAPOR
also has procedures for registering complaints against pollsters who
violate its code of professional ethics and standards. There are
self-correction mechanisms to deal with the sort of over-the-top fraud
you describe, even if it isn't raised to the level of criminality.
The reputable pollsters have as much interest as anyone in seeing the
push-polls, salesmen-masquerading-as-pollsters, etc. removed from the
scene, because the bad apples make the entire industry suspect.
You should also note that the days of live interviewers with clipboards
walking down the street and taking a poll are pretty much gone, so the
sort of biasing sample by physical appearance you describe isn't much of
an issue. The vast majority of election or other public opinion polling
is done via telephone (usually RDD, or Random Digit Dial, although some
use RBS, or Registration Based Sampling), with various means of
screening respondents. It's possible, I suppose, that an election poll
could have "do you have any body piercings" and use that as a screening
question, but it seems a bit silly to do so when you can just as easily
ask "Are you a Republican or a Democrat?", or "Are you conservative or
liberal?" and weight the sample accordingly.
More information about the extropy-chat