Up and Down: The inexact science of polling

uring a campaign season, it is impossible to pick up a newspaper, turn on a television or radio, or surf the Internet without seeing or hearing the latest poll. Polls are a fact of life in politics, but how much do we know about polls and how much should we trust their results? The follow article gives background on polls and tips on how to be good consumers of their findings.

Can polls be trusted in the first place? And if they are reliable indicators of public opinion, why do the vary so much? Polling is still an inexact science subject to error as well as manipulation. Yet, when they are conducted properly, polls produce a reasonable approximation of where an election race stands at a given time.

Time, however, is often the key wild card. People's opinions may change often, particularly if they lack a strong bond to a party or candidate. Polls can also vary because of the way a sample is chosen, the wording and order of questions, even the time of day respondents are contacted.

"You have to look at several polls over a period of time and average them," says Karlyn Bowman, polling analyst at the American Enterprise Institute. That's what campaign strategists do when tracking the media polls and that's what they say voters should do as well.

Historically, presidential polls have had a remarkably accurate track record of predicting the winners: The leader on Labor Day is the victor in November. The only recent exception was in 1980, when many polls showed Reagan and President Carter virtually tied in early September. Reagan went on to win the election by 10 points.

Like them or not, polls are a fixture of the political process, USA Today Executive Editor Bob Dubill says. "People are curious about election campaigns. Everyone--journalists, politicians, campaign operatives and the public at large--wants to know who is leading, what's changed and why. Reporters can't interview everybody to find out. But scientifically drawn and conducted polls can take the pulse of the public and provide a sharp snapshot of where the race is at any given time."

Beyond the "horse race" question that gets most of the attention, media polls also gauge the mood of voters and their stand on issues. Dubill adds, "Journalistically, we would be remiss if we didn't sample the public attitudes," he says. "Polls are the best way we can allow the public to have its say. In a democracy, polls are news."

Then President Truman displays a Chicago Tribune headline that, based on polls, incorrectly predicted the outcome of the 1948 Presidential election. See story at right form the Burlington Free Press, November 4, 1948.

Evaluate The Polls

Why do two polls taken over the same time period produce different results? And why can surveys by the same polling organization fluctuate so much within a week or two? Pollsters point to several factors that affect results:

How the sample is selected. The sample is supposed to be selected on a random basis with everyone theoretically having equal change of being picked. The sample also should be representative of the population as a whole.

How big the sample is. Generally, the bigger the sample size, the more likely the results accurately capture public attitudes. A good national sample should have at least 1,000 adults.
Who is included in the sample. Is the poll surveying all adults, registered voters or likely voters? The best measure is likely voters.

How the questions are worded. The wording of questions can affect an answer. For example, if a poll asks if you agree or disagree with a plan to reduce aid for textbooks for school children, few would agree. But if the poll asks if you agree or disagree with a plan to shift money from textbooks to computers, the response might be quite different.

How questions are ordered. Some polling organizations always ask the gubernatorial horse-race question first so that other questions can't influence the respondent's reply.

When the poll is conducted. The most reliable polls are taken over several days. Polls taken during the week generally produce more representative samples than those conducted over weekends when many people are not at home.

What news events are occurring during the polling. Voters' responses often are based on the last thing they heard reported about the campaigns. That explains in part why gubernational candidates usually get a boost after campaign kick-off events.

What is the margin of error. The margin of error is a measurement of how likely a poll reflects actual public opinion. If a poll has an error margin or +/- 5 percentage points, it means the actual support for a candidate favored by 47% of those surveyed could be as high as 52% or as low as 42%, a 10-point range. The margin of error shrinks as the sample size grows.

DID YOU KNOW?
Many candidates and elected officials use polls to help gauge public opinion on policy issues?

ASK THE READER
Should candidates and elected officials use polls to help determine policy positions and planning?

TALK BACK
Here are the results of last week's question "Should the voting age be raised to 25?"

86% voted NO 14% voted YES

Here's some feedback:

"If 18-year-olds can marry and own a gun, they should be able to vote."

"I am 12 years old and I follow the election as closely as my parents. Children have opinions, too. To raise the voting age would be saying we don't trust you to make a good decision."