Throughout the election season, I have received many questions like
the one that Paul V. Bass asked Joseph Farah recently. Mr. Bass
wondered, “Why is it that all the polls shown on TV show Al Gore as the
front runner, yet every poll I have ever participated in online, clearly
shows George W. Bush as having a large lead. How come these polls aren’t
brought to the attention of TV news personnel? Also, most of the polls
reported on the TV news or in newspapers show maybe a thousand or less
people being polled, the polls on WND or even AOL show many thousands of
people supporting Bush?”
Before addressing the central question raised here, let me point out
that Fox News regularly reports our surveys and the Battleground surveys
which have both shown Bush in the lead since Labor Day.
Now, let’s begin with the first question raised by Mr. Bass.
The Internet, while a great source for news and data, is a terrible
place for conducting accurate polls. For a poll to be reliable it must
use a random sample of respondents. Internet polls are self-selected,
not randomly selected. In other words, people who participate in
Internet polls choose to do so. They are not contacted in the random and
independent manner used for telephone surveys.
To understand the importance of this, imagine the poll results you
might receive from an on-line poll whose participants are the
WorldNetDaily audience. It would probably lean heavily to conservative
and libertarian views. In the Presidential election, the WND audience
would lean heavily towards Bush. Not only that, Harry Browne would
probably do better than Ralph Nader, and, in the WND online polls, that
is exactly how it’s shaken out thus far.
Now, imagine the same poll conducted with participants selected from
Hillary.org. That poll would probably show more statist and liberal
leanings. In the presidential race, Gore would clearly be in the lead
and Harry Browne might not get any votes. Same survey, same technology,
The point is that both polls would, to some degree, reflect the
audience visiting the site hosting the poll. Neither poll would reflect
the public at large or the likely outcome of the Presidential election
this year. The only way to get a true representative sample of the
population is to randomly select your survey participants.
A great example of this phenomenon was discovered one month ago when
Rasmussen Research asked 822 randomly-selected Americans to take “The
World’s Smallest Political Quiz.” The quiz, a product of the libertarian
leaning Advocates for Self-Government, is posted on the Advocates’
website and asks participants 10 questions that position them on a
liberal, conservative, centrist, authoritarian and libertarian scale.
Our telephone survey found that 16 percent of Americans qualify as
libertarian according to the test. However, in the online survey
conducted on the Advocates for Self-Government site, 38 percent
qualified as libertarians. Again, if this exact same test had been
conducted on some other website, the number would have been entirely
The most logical explanation for this result is that visitors to the
Advocates’ site are far more likely to be libertarian than the
population at large. Our telephone survey reached many people who do not
use the Internet, do not know what a libertarian philosophy is, and have
never heard of the World’s Smallest Political Quiz. This sample was far
more reflective of the population at large.
So, the biggest problem with Internet polls is the fact that
participants are self-selected.
There’s another major problem as well: not everybody uses the
Internet. Even if an Internet poll draws respondents from many different
states, ethnic backgrounds and ages, only about half the U.S. population
is online. So, even if you could magically come up with a way to use
on-line polling to accurately measure the views of the online community,
there is no way such polls can take into account the views of
As far as sample size is concerned, the second half of Paul’s
question, that will be a topic for next week’s column.