Public Opinion

Politicians have long been interested in the moods and opinions of the people. After all, members of Congress and the President are elected to represent the people, a difficult task if one does not know and understand what the people want. Even George Washington employed his own "pollster," a friend back in Virginia who would mingle with the "ordinary folks" to find out what they thought of the President.1 While the measurement of public opinion has become much more scientific and precise since Washington's day, the attitudes of the public have always played an important role in shaping public policy and the direction of the nation.

What is Public Opinion?

"Public opinion" is the distribution of opinions and attitudes held by the public. Individuals hold a wide variety of opinions. By measuring these opinions at the individual level and aggregating them, the proportions of the population with particular beliefs and preferences can be determined. When measuring public opinion, pollsters are not only concerned with the content of public opinion, i.e. what it is that people think and believe, but also with the stability of people's opinions over time. It is also useful to know how strongly the public holds particular opinions and the direction those opinions seem to be moving. The content, stability, intensity and direction of public opinion are all important indicators of what the public wants and expects out of its government and political leaders. While popular opinion is not the only factor that determines public policy, it is generally the most important one. The role of public opinion in a representative democracy is further discussed below.

Measuring Public Opinion

Public opinion is not measured by asking every individual in the population questions about his or her opinions. In a nation of 270 million people, doing so would be impossible. Instead, pollsters ask a small group of individuals a set of questions and then generalize from their findings about that group to the larger population. Choosing a group of individuals out of a population that is representative of the entire population and then accurately measuring the opinions of those individuals is technical and complex.

Sampling

A group of people selected out of a population to be asked questions on a survey is called a "sample." For the results of a survey to be accurate, the sample must be similar to the broader population in terms of race, age, sex, education, income and other important characteristics. If the sample is not representative of the population, the opinions of those in the sample cannot be generalized to the larger group.

How big does a sample need to be? The answer might surprise you. First, sample size depends much more on the amount of diversity in a population than it does the size of that population. For example, if every individual in a one million person population holds the exact same opinions, a sample size of one would be sufficient to accurately gauge the opinions of the group. In contrast, if every person in a group of one thousand people held very different opinions on a subject or subjects, a sample size of one thousand might be required to accurately measure the group's attitudes. (Technically, if every member of a population is surveyed, there is no "sample" or "sampling" involved, just the population.) A sample must be large enough, then, to reflect all of the significant variations in opinions and attitudes throughout the population. So how big is big enough? For small populations, say of five hundred or less, a researcher is generally better off simply surveying the entire group. In contrast, a sample of five hundred is generally sufficient to measure the opinions of a population of 100,000 people. In fact, as population size increases, the proportion of a population required to comprise an acceptable sample decreases. Repeated experimentation with varying sample sizes and extensive statistical and probability analysis has led to the conclusion that a relatively small, randomly drawn sample of 1,200 or 1,500 individuals can be highly representative of a population of more than 250 million people. In fact, increasing sample sizes to 5,000 or even 100,000 does not substantially improve the accuracy of the results.

In theory, selecting a random sample is straightforward. First, a researcher must decide what group of people he or she is interested in learning about. For example, if someone is interested in predicting how people will vote in an upcoming election, the relevant population consists of those individuals who are going to vote. Similarly, if one wants to measure the opinions of homeowners, the population consists of people who own homes. Once a population is defined, the next task is to select a representative sample. The most effective way to do this is to randomly select individuals from a list of every individual in the population. In most cases, however, it is difficult or impossible to obtain a list of everyone in the relevant population. For instance, in the previous voting example, the list of those who actually vote will not be determined until election day. In such cases, researchers must do their best to randomly select people who are highly likely to be in the population.

In practice, most public opinion surveys are conducted by telephone. Using specialized computer software, random telephone numbers are dialed and interviewers attempt to complete interviews with persons in those households. To promote the randomness of the sampling process, interviewers may also be instructed by the computer program to ask for either a male or female respondent. Another less-frequently used sampling technique involves randomly selecting addresses in a neighborhood or area and then having interviewers personally visit the homes at those addresses. A more commonly used but more problematic technique is to mail surveys to randomly selected individuals or households. 

Scientific v. Nonscientific Sampling

When a random, representative sample is drawn from a population, the procedure is deemed "scientific" sampling. When the procedures of scientific sampling are not followed, however, the people in the sample are very unlikely to be representative of the broader population. For example, Washington's informal "mingling" polls mentioned above did not produce samples that reflected the general opinions and views of the people. Similarly, the "call-in-your-vote" polls that often appear on television news programs or the "instant" polls conducted on Internet web sites do not draw randomly selected individuals. Indeed, these polls are almost completely worthless in terms of their generalizability to any recognizable larger population.

One of the most famous public opinion polling blunders resulted from relying on a sample that was not representative of the population. Several days before the 1948 presidential election, pollsters predicted that Thomas Dewey would defeat Harry Truman by as much as fifteen percentage points. The Chicago Tribune was so confident in the prediction that it ran the headline "Dewey Defeats Truman" on the front page of an election night newspaper. (In the photo above, Truman is showing the paper to a crowd of supporters.) It was Truman, though, not Dewey, who won by 4.5 percent. What had gone wrong? The errors in prediction were at least partly due to poor sampling--the pollsters had asked all eligible voters who they would vote for on election day. Truman supporters turned out to vote in much larger numbers than Dewey supporters, however, making the voting population very different from the samples the pollsters had been questioning.

Despite some early setbacks (see the box on sampling above), pollsters have become very skilled at drawing representative samples. However, the single most difficult sampling problem is not drawing a representative sample, but rather securing completed interviews with each selected individual. For a variety of reasons, many of the people selected to be in a sample will not answer the questions on a survey--they may be too busy, too disinterested, or they may simply be unavailable. If the people who refuse to complete the survey share similar characteristics, such as working at the same time of day or having similarly hostile attitudes toward public opinion surveys, their absence from the sample will diminish the survey's generalizability. Unless all kinds of people are included in the sample, even those who don't like to answer surveys, the sample will not be a representative cross-section of the entire population.

Asking Questions

Assuming that every member of a random sample is willing and available to answer the questions on a survey, pollsters must still design valid and reliable questions and then ask them of each individual in the sample. If the questions that are asked are not carefully written and presented consistently, the results may be inaccurate even with a good sample. Several problems associated with survey questionnaire design and interviewing must be anticipated and accounted for to make the results as accurate as possible.

Pollsters must exercise caution in writing questions because the way a question is worded can influence how individuals respond to it. Leading questions, loaded phrases, poorly worded questions, providing responses is a particular order or even the order in which questions are asked can all effect the responses that are given.

For example, if a researcher was interested in measuring public support for a state governor or member of Congress, he or she might simply ask: "Do you approve or disapprove of the way [insert the politician's name] is handling his/her job as [insert title of office]?" Altering this question in any way might change the results of the survey. Pollsters have discovered that even including titles such as "Governor," "Senator," "Congressman," or "President" tends to slightly increase the approval ratings of politicians. Using descriptive words, such as "successful" or "hard-working," can alter the results even more. Asking the question as it stands immediately after asking a question that casts the individual in potentially positive or negative light can also alter individual responses. For example, if a researcher asked respondents what they would think of a politician if they found out he or she was having an affair and then asked the "approve or disapprove" question, the results would be less favorable for the politician than if the question had been asked in isolation.

The accuracy of individual responses to survey questions can also be influenced by unwillingness (because of evasiveness) or inability (because of a lack of information or poor memory) to answer questions accurately. Respondents may also give "socially desirable" answers instead of accurate ones when asked about such positively viewed activities as voting or donating to charities or negatively viewed activities such as criminal behavior or discrimination. The appearance and tone of voice of the interviewer can also influence the way people respond to survey questions.

How to Read Poll Information

Despite all of the difficulties associated with accurately measuring public opinion, skilled and conscientious survey researchers are still able to conduct public opinion polls that yield valid and reliable results. As a consumer of the news and of public opinion polls, you should, however, look at public opinion poll results with a critical eye. Was the sample selected randomly? What was the response rate of individuals in the sample? What kinds of questions were asked? In what order?

In most cases, you will not be provided with enough information to answer all of these questions. A growing number of news outlets, however, are using the Internet to disseminate detailed information about the surveys they conduct, often publishing the entire survey questionnaire online for readers to see for themselves. In almost every case though, what is reported about a poll is the sample size, the dates the poll was conducted and the "margin of error" for the poll.

 

Survey A
Conducted
9/1

Survey B
Conducted
10/1
 
n=1200
+/- 5%

n=1500
+/- 4%

Candidate
Fred
55%
40%
Candidate
Wilma
45%
60%

A poll's "margin of error" is a statistical statement of the amount of confidence the survey researcher has in the results of his or her poll. Most public opinion polls have a margin of error between plus or minus seven and plus or minus three percentage points. What this means is that the percentages of individuals in the sample who chose each response to any given question could be off either direction by the magnitude of the reported margin of error. 

In the example to the right, the results of Survey A show that Candidate Fred is supported by 55% of the voters while Candidate Wilma is supported by 45%. The margin of error for Survey A is plus or minus 5%. This means that each candidate's support could be as much as five percentage points higher or lower in the population than it was measured in the sample. In this case, adding five points to Candidate Wilma's total and correspondingly subtracting five points from Candidate Fred's would put each candidate's support at 50%. Survey B, taken a month after Survey A, shows that Candidate Wilma is now supported by 60% of the voters while Candidate Fred has dropped to 40%. With a margin of error of plus or minus 4%, the lowest Wilma's support in the population is likely to be is 56%, while Fred's likely high is 44%. Even adjusting for the margin of error in Survey B, Wilma enjoys a clear lead over Fred. Based only on the results of Survey A, however, the race is too close to call because the difference between the two candidates is within the margin of error.

Public Opinion and Democracy

As was noted at the outset, public opinion has always been an important influence in American politics. Indeed, even the Federalists sought to build public support for the Constitution during the debates on ratification so the delegates to state ratifying conventions would feel compelled to cast votes in support of the document. That is not to say, however, that the Federalists trusted the opinions of the people. On the contrary, the representative system created by the Constitution reflects their fears about government by majority opinion. In particular, the Senate, with its members originally chosen by state legislators, not directly by the people, was designed to insulate public policy decisions from the opinions of the people. Indeed, James Madison declared in The Federalist No. 63:

[The Senate] may be sometimes necessary as a defense to the people against their own temporary errors and delusions. As the cool and deliberate sense of the community ought, in all governments, and actually will, in all free governments, ultimately prevail over the views of its rulers; so there are particular moments in public affairs when the people, stimulated by some irregular passion, or some illicit advantage, or misled by the artful misrepresentations of interested men, may call for measures which they themselves will afterwards be the most ready to lament and condemn. In these critical moments, how salutary will be the interference of some temperate and respectable body of citizens, in order to check the misguided career, and to suspend the blow meditated by the people against themselves, until reason, justice, and truth can regain their authority over the public mind?

In no uncertain terms, the Framers of the Constitution anticipated that the opinions of the people would influence the decisions made by political leaders, but they also hoped that the representatives of the people would not simply mimic the passions of the majority at any given moment. Indeed, the process of representation, as it is practiced by most members of Congress, is a much more careful and considered exercise than reading opinion polls and voting accordingly (see "Congressional Representation").

Should the opinions of the people be trusted? An abundance of survey evidence suggests that the people, on the whole, display an astonishing lack of information about the political system, its processes and about particular policy problems and issues. If public opinion is so poorly informed, would it not be easily manipulated? That is at least one reason the Framers favored republican, i.e. representative, government over democracy. By requiring the representatives of the people to periodically face reelection, they provided a mechanism for keeping them accountable for their decisions. In between elections, however, members of Congress could cast votes in Washington and then go back home and explain them. In a republic, there are bound to be differences of opinion between the people and their elected representatives. The Framers appear to have hoped such differences would arise regularly. When such differences exist, though, it is the responsibility of each representative to explain to the people back home why he or she has cast a vote that was contrary to the will of the people. If the explanation is inadequate or unconvincing, the people can choose to send a new representative to Washington.