|
In
the weeks leading up to election day, it is not unusual to see polls
disagree about which candidate is leading the race and by how much.
But if it seems like they have been even more contradictory in this
cliffhanger of a presidential contest, the people who make their
living taking the public’s pulse say there are some obvious
reasons why.
"I think the differences we’re seeing in presidential
poll results are generally due to real changes in public opinion,"
said San Francisco pollster David Binder. "The Bush campaign
has advertised in California while the Gore campaign has not, and
this has been reflected in the results."
But genuine movement in public opinion is not the only factor that
can cause contradictions and sometimes confounding errors in the
very scientific process of public polling. Binder described three
important factors to consider when looking at changes in the polls.
"Turnout assumptions by different pollsters vary widely. Who
the pollsters think will come out to vote and how they base their
polling responses upon this assumption has a definite impact on
overall results," he said.
Sample criteria such as random digit telephone dialing or calls
based on voter registration, depending on the sample size, could
also introduce error. How questions are phrased is a major aspect
of polling and is the responsibility of the pollster hired to conduct
the poll. Neutral question wording is an intangible skill, often
based on experience and technique, and pollsters share information
from polls conducted. Even where a question is placed in the interview
can have a drastic effect on results.
"If the voter is warmed up to the question ‘Who are you
going to vote for for president’ rather than asked right from
the start, there is a likelihood their answer may be affected,"
said Binder. "If there are previous questions that cause them
to think about the issues, their opinion on the economy, on abortion
or on taxes, then their answer for who they will vote for president
will be much more honest."
But to academics like Henry Brady, Professor of Political Science
at U.C. Berkeley, there is another reason poll results vary, one
that most political pollsters don’t like to talk about.
"The response rate is not described in many of these polls,"
said Brady. "I have a standing bet for $100 dollars to anyone
who can find me a poll that lists response rates along with results."
To understand Brady’s skepticism, it is important to know that
not all polls are designed to measure the same thing. Candidates
typically conduct a baseline poll at the start of the campaign,
a major undertaking giving them a likely idea of how their constituents
view the issues. The purpose is not to predict who will win the
election, but rather to determine what positions or strategies might
push voters one way or the other.
These baseline polls can cost up to $35,000 and take on average
about 20 minutes per call. The typical sample is 400 people, but
this number can be increased if the candidate wishes to understand
precise results from a specific group of voters.
Tracking polls, usually costing around $5,000 and taking only about
four minutes to conduct, are used to monitor the progress of the
candidate or his commercials and mailings.
Brady did not feel it possible that every demographic could be reached
for their opinions if tracking polls are all conducted within one
night. Brady said, "Tracking polls are usually based on only
one phone call, and poor people or the elderly or minority groups
are not always at home."
Although Binder somewhat agrees with this statement he also feels
that it is the pollsters responsibility to take factors such as
this into consideration when conducting a poll. "The cardinal
rule for pollsters is to draw a sample that is reflective of the
likely voter universe," he said.
If random sampling techniques are applied, Binder says, then results
should be balanced out in the long run. "The quality of the
pollster is reflected in the quality of the sample, and a bad sample
equals a bad pollster."
|
