Polling for Truth

This is a time of debate and introspection. Last week, we had to go to class, and choose which professors would determine the course of our next four months. In seven weeks, America will go to the polls and decide who the leader of the free world will be.

Today, accuracy is at a premium.

In this crucial moment, more than any other, we depend upon our media to do their homework, if we are going to have to do our duty.

Yet, the press fails us.

The media are frequently misrepresenting the meaning of polls. In a Sept. 19 article on the state of the presidential election, "Polls have mixed Results for Gore," Associated Press writer William Lester stated that, "the CNN/Gallup poll gives Gore a 5 percent edge, while the Voter.com poll puts Gore and Bush in a dead heat." He implied that while one poll shows the candidates tied, the other shows Gore to be ahead. This isn't true.

A poll shows that a candidate is ahead only when the difference between his support and the support for another candidate is twice the margin of error.

Polls measure averages. The margin of error demonstrates the range where that average could be. For example, if 45 percent of voters say they will vote for you, in a poll with a margin of error of five percent, that means your support is somewhere between 40 and 50 percent. If in the same poll, your opponent gets 39 percent support, the six percent difference does not mean you are ahead. Because your support could be as low as 41 percent, and that of your opponent as high as 44 percent, you are tied.

The Gallup poll (where Gore gets five percent more support than Bush) has a margin of error of four percent. The Voter.com poll (where Bush gets two percent more support than Gore) has a margin of error of three percent. Thus, in both polls, Bush and Gore are in a "statistical dead heat." No one has an edge. Nevertheless, this article, like many others I have read, mistakenly puts Gore ahead.

Newsweek has been the most blatantly inaccurate, giving Gore six to 10 percent post-convention leads in August and early September, then taking several pages to explain why he's so popular. Three things should make people question the credibility of these polls. First, the results of these polls have generally shown more support for Gore than most others, including Reuters, Gallup, and the Battleground poll. Second, the headlines are based solely on registered voters--a fact that might be easy to overlook, given that it took five or more paragraphs to discover. Third, Newsweek has lauded Gore with such headlines as, "Candidate's speech propel him to a lead," (August 19th) when the margin of error indicated a tie, demonstrating either an ignorance of what their polls mean, or a reluctance to speak about it.

Who is polled is probably the most important factor in evaluating the credibility of what polls have to say. According to Gallup, registered voters, the group Newsweek profiles most prominently, tend to be more Democratic than those who actually vote. Not surprisingly, the Newsweek survey leans to the left of more proven polls.

Over the last two weeks, Newsweek has begun to mention the results among likely voters. But without mentioning the margin of error among them, the polls remain meaningless.

I should point out one other detail. Seeing the same poll results each morning shouldn't reinforce people's beliefs about the strength of a candidate's lead. It turns out that the most cited polls, those coming out each day from Voter.com and Gallup, are rolling surveys. They represent three to four day trends. When a new poll comes out, it's actually 67 to 75 percent an old poll. So, not surprisingly, the polls don't change much day to day.

Seeing the same results over a short period could cause people to think, incorrectly, that one candidate is more ineffective, or resilient, than he actually is. If a candidate comes out with a major new initiative, as Bush did recently with his "Real Plans for Real People", the effect of the initiative won't appear for about a week. This isn't because the candidate's message isn't working. Its simply because of the way polls are constructed.

I hope public misperception and media ignorance of statistics don't adversely influence the electorate's perceptions, and ultimately, erroneously determine our choice in November.

Beau A.J. Briese is an undergraduate in Kirkland House.