Public opinion poll is a survey that attempts to find out the attitudes, beliefs, or opinions of a target population. A target population is a particular group of people that is identified as the intended recipient of an advertisement , product, or campaign. The population covered by a poll may include millions of individuals, but only a small number of them are actually questioned. If the people who are questioned have been properly chosen, their opinions will usually reflect those of the entire group. Public opinion poll respondents are usually randomly selected. Public opinion polls may be conducted through personal interviews, by telephone , by mail , or over the Internet . People who conduct polls are sometimes called pollsters.
Public opinion polls ask a wide variety of questions. For example, a poll before an election may ask people whom they plan to vote for and why. Other polls may ask people if they plan to buy a new car, how they feel about their work, or whether they use certain government services. Polls are conducted throughout the world, but they are most frequently used in developed countries with democratic governments.
Who uses polls?
Public opinion polls are used chiefly by five types of groups: (1) news media, (2) politicians and political organizations, (3) business companies, (4) government agencies, and (5) social scientists. These groups generally use polls that are conducted by private pollsters, university research centers, or government agencies. Some groups conduct their own polls.
News media
—including newspapers , magazines , television , websites , and radio —publish or broadcast the findings of public opinion polls. Many major television networks and newspapers have internal public opinion polling operations and conduct their own polls. Broadcasting stations and newspapers may subscribe to polls from such polling companies as the Gallup Organization, Harris Interactive , or Pew Research.
Politicians and political organizations
use polls to help them plan their election campaigns and political lobbying efforts. They also use polls to keep track of how the public feels about political leaders, the economy, and other issues of public concern. Polls can help elected officials make decisions by telling them how people feel about various policies and issues. Many political candidates and elected officials hire polling organizations to conduct private polls for their internal use.
Business companies
use polls to help them manage their operations and sell their products. Many businesses also study the polls that appear in the news media. Some companies subscribe to special polls taken by private polling companies. Many advertising agencies conduct market research polls that measure people’s opinions of a product or reactions to an ad campaign.
Government agencies
rely on polls for guidance in operating and evaluating their programs and in identifying areas of public need. Such polls may ask people’s opinions on educational programs, medical services, transportation, and other subjects.
Social scientists
sometimes use polls when studying human behavior . A psychologist , for instance, might conduct a poll among different age groups to study differences in attitudes between younger and older generations. A political scientist might conduct a poll to study how citizens learn about politics. A sociologist might conduct a poll to learn how people think about race, gender, or religion.
Conducting a poll
The work of pollsters involves five main steps: (1) defining the goals, (2) selecting the sample (the individuals to be questioned), (3) designing the questionnaire, (4) interviewing the sample, and (5) analyzing the results.
Defining the goals
involves deciding what a poll will seek to find out. A poll may ask people’s opinions about certain economic, political, or social issues. It may study people’s attitudes toward events, individuals, or situations. The group of people that a poll seeks to generalize to (learn about) is called the population or universe. A population may consist of everyone in a city, state, country, or some other area. In some cases, it may include only a certain group, such as factory workers, homeowners, teenagers, or users of a specific program or service , such as students in a school district or social service recipients.
Selecting the sample.
Pollsters try to select the sample so that every person in the population has close to an equal chance of being contacted. In polls conducted through personal interviews, pollsters will often stratify (divide) the area to be surveyed into geographic regions so that the poll has sufficient geographic coverage. Specific localities are then selected by chance. Within these localities, the pollsters randomly select a certain number of neighborhoods or census blocks. A census block is the smallest geographic unit used by the U.S. Census Bureau . Pollsters then conduct interviews in several households in each of these neighborhoods or census blocks.
When polls are conducted by telephone, pollsters usually select the sample through the use of random digit dialing. In this technique, phone numbers are randomly selected from among all possible home phone numbers in the geographical area being polled. In polls conducted by mail, questionnaires are sent to randomly selected household addresses. However, in mail surveys, people in the selected sample sometimes do not return the questionnaires. As a result, mail surveys are usually not as reliable as other forms of polling.
In some Internet polls, people are randomly selected for questioning. In other cases, the pollster assembles a panel (large list of individuals who say they are willing to take polls) and the pollster randomly selects names from the panel. In still other cases, the surveys are open to any Internet user who wishes to participate. There is active debate about the reliability of Internet polls. Some pollsters believe that by using a statistical technique called weighting, an Internet poll can be as reliable as other kinds of polls. Weighting is assigning a factor to a number in a calcuation to make the number’s effect on the calculation reflect its importance. Other Internet polls, particularly those that are open to any participant, often have difficulty establishing a sample that accurately represents the population. However, because Internet polls cost far less than personal interviews and obtain higher response rates than telephone polls, they are used more and more frequently for all kinds of polling.
Designing the questionnaire.
Pollsters ask two general types of questions: closed and open. A closed question asks the respondents to select their answers from two or more choices. An open question asks them to give their opinions in their own words.
Before pollsters conduct a poll, they may assemble a focus group of about 10 people to discuss the issue the poll will examine. The group’s comments help pollsters understand the public’s attitudes toward the issue so that they can construct an effective questionnaire. Most pollsters pilot (pretest) the questionnaire on a small number of people before starting the actual poll. By pretesting, the pollsters can tell if the respondents understand the questions, and if the answers provide the information sought. They also find out if the order of the questions affects the way people answer them.
Interviewing the sample.
Most pollsters question respondents in person or by telephone, or ask people to submit their answers by mail or over the Internet. Personal and telephone interviews have traditionally been thought to be the best methods for ensuring that every person, or almost every person, in the selected sample is questioned. Telephone questioning is less expensive than personal interviewing. However, the number of people willing to speak with an interviewer has fallen dramatically in recent years. Other people no longer maintain a home phone line. Personal interviews have other advantages as well. For instance, a personal interviewer can be reasonably sure that the respondent understands the questions. A personal interviewer can also use cards or other displays that list choices of possible answers.
Internet polling has been growing in popularity for private, academic, and governmental polls because it is much less expensive than personal interviews and its response rates are higher than those of telephone polls. Some Internet polls allow respondents to view photographs or video presentations to assist with the questioning process. Internet polls can also be used to conduct survey experiments in which respondents are asked various questions as a way to help learn more about public opinion.
Analyzing the results.
The most common tabulation of responses shows the percentage of respondents who answered each question in a certain way. Analysis of the results can show how strongly people feel about various subjects and whether their opinions have changed since a previous poll. It can also show what differences of opinion exist between different segments of the population and how attitudes on different subjects are interrelated. Complex statistical analysis can help pollsters understand the reasons for many poll results.
Evaluating a poll
The reliability of a poll depends chiefly on the size of the sample and how it is drawn. Most national polls involve interviews with between 500 and 2,000 people, depending on the purpose of the poll. If scientific procedures are followed in selecting the sample, the pollster can calculate sampling error. Sampling error is expressed as a certain number of percentage points or a certain range above and below the reported finding of the poll. These estimates are the range of results that could be expected if a large number of similar samples were asked the same question. Sampling error depends on the size of the sample, not the size of the population. For example, a 1,000-person poll is just as reliable whether the target population is 10,000 or 100,000,000.
Questions that are not fairly worded can also affect a poll’s reliability. In addition, polls that have been sponsored by individuals who have something to gain by a certain result should be regarded cautiously.
History
A Pennsylvania newspaper conducted one of the earliest surveys of public opinion in the United States. In 1824, the Harrisburg Pennsylvanian asked voters in Wilmington , Delaware, who they thought would be elected president that year. On the basis of the poll, the newspaper predicted that Andrew Jackson would win. Jackson received more votes than any of his three opponents, but he did not get a majority. As a result, the election went to the House of Representatives , which elected John Quincy Adams .
Polls following scientific procedures were first used in 1935 with the experimental nationwide surveys of George H. Gallup and Elmo Roper. Another pollster, Archibald M. Crossley, began using scientific polling methods the following year. In 1940, the first academic center for the development of polling techniques was established by Hadley Cantril at Princeton University.
The modern era of scientific polling began in 1936, after a controversial poll. That year, the magazine Literary Digest mailed out 10 million questionnaires concerning that year’s presidential election. Two million questionnaires were returned. Based on these replies, the magazine predicted that Governor Alfred M. Landon of Kansas would defeat President Franklin D. Roosevelt . But Roosevelt won the election by a landslide, an outcome that George Gallup was able to predict using a much smaller, randomly selected sample. The Literary Digest poll was inaccurate chiefly because the questionnaires were mailed to people chosen from telephone directories and from lists of car owners. As a result, wealthy people were overrepresented in the sample. The controversy established in the public mind the scientific validity of randomly sampled polls.
In 1948, polls based on quota samples—that is, samples with a certain number of people from a variety of categories—predicted that Governor Thomas E. Dewey of New York would defeat President Harry S. Truman . However, Truman won the election. This time, Gallup, along with Roper and Crossley, all mispredicted the winner, based on their polls. The polls failed for two chief reasons. The last polls taken were conducted too long before the election, and many voters changed their mind. Also, the quota samples did not accurately represent the people who voted. After the 1948 election, most pollsters began to use random sampling. This change, along with refinements in interviewing and other procedures, increased the reliability of polls.
Since the 1970’s, two special kinds of polls, called tracking polls and exit polls, have been widely used during elections. Tracking polls are conducted with small samples during the course of an election campaign. Candidates use these polls to follow changes in their standing with voters. Exit polls are taken as people leave their voting place. The news media use exit polls to help interpret the results of elections.
