Measuring Public Opinion

Exploros_logo

Measuring Public Opinion

In 1932, George Gallup’s mother-in-law ran for public office in Iowa. She was running against a popular incumbent, and no one expected her to win, except her son-in-law. He polled her constituency, told her she could win, and gave her some advice. Hers was only the first of many elections he predicted correctly.

Polling organizations now play an important role in American politics, measuring public opinion with their precise methods.

Polls try to answer a political question. For example, "How many people in my district know who I am and what I do?" or "How do people in this country feel about big tobacco companies?"

The pollsters follow these steps:

Questions must be carefully and objectively worded. The slightest shift in the wording of a question can bring very different results.
The sample must be randomly selected. First, pollsters determine the group whose attitudes they wish to measure. Since they cannot question everyone, they use random sampling, which gives each potential member of the group the same chance of being selected.
Respondents must be contacted in a cost-efficient way while maintaining accuracy. For example, asking television viewers to call in their opinions is generally not very accurate, because the people that call in usually feel very strongly about the issue, and some of them call in more than once.

Should politicians monitor the polls? Candidates have been criticized for shifting their positions based on the results of public opinion polls. But a politician is supposed to represent the true will of the people.

Poll results must be carefully and accurately compiled and reported. This is not always an easy task, especially for tracking polls that measure changing public opinion. A good example is an election poll. Statistics that are a week old are unreliable when trying to predict a close presidential race.

Polls can never be completely accurate because a sample doesn't fully represent the entire group. Pollsters allow for this slight chance of inaccuracy with a margin of error. Standard samples of about 1,000 to 1,500 individuals can usually represent a group of millions of people with only a small amount of error. A typical margin of error is about 3%.