"Jim Newell" wrote in message
...
I don't know to be honest with you. Often, it seems that the polls trend
in
a certian direction in politics, but the facts do not match expectations
of
the poll. One can't help but wonder if there is not some bias at work.

Many people think surveys are biased, Jim. The fair response to that would
be the contents of a good program on the methods, uses, and misuses of
surveys. It doesn't fit into an online message.

Maybe I can plant one thought: The big, reputable polling firms, including
Harris, Gallup, Pew, and so on, make most of their livings doing private
research for businesses and other institutions that want the best, most
accurate statistical information they can get. The results of such research
is rarely made public. It isn't for the purpose of propaganda; it's for
making business or policy decisions, privately and quietly.

If one of those research firms got caught playing games with their publicly
available research, they'd be out of business in a heartbeat. Their
integrity, and their scientific skill at running such surveys, are all they
have to sell.

One thing we run into frequently, however, is well-done surveys that ask
inherently biased questions. Don't blame the survey firms for that. They're
contracted to get an answer to *that* question. Any misuse of the results is
the fault of the institution contracting for the survey. If it was a
legitimate polling firm that did the survey (and it usually is), then you
can uncover such biases in a few minutes by tracking down the question(s)
asked. Legit pollsters publish the actual questions and the methodology.
Anybody with a clear head will recognize a question that prompts a specific
answer, or that was misrepresented in the reporting of the results by the
institution that had the survey done.

In this case, if you followed the link I posted to those summaries of over
20 surveys, you'd see that the question was asked similarly in almost every
one: "And if the election for Congress were held today, would you want to
see the Republicans or Democrats win control of Congress?," or something
similar to that.

When you get that many surveys, by many different polling organizations,
under contract to a variety of political interests, asking the same
question, who come up with a consistent qualitative answer...it's time to
pay attention. In fact, if one survey had been done, it would be worth
paying attention. It just would require more work, and more knowledge of
statistics and survey methods, to form a strong opinion.

About your comment regarding the facts matching the polls, they almost
always match the results of the poll. You just have to be careful what
you're reading, and you have to make sure you aren't reading something into
it that isn't there. It's also a good idea not to trust one's memory. We
tend to be a lot more biased in individual memory than in survey statistics.
And watch out for reporters' interpretations of surveys. Most of them
wouldn't know a standard deviation from a confidence interval, and they have
no business expounding upon survey research results.

And remember: Statistics means never having to say you're certain. g

--
Ed Huntress