This booklet includes the study of public opinion is relatively new but the practical study of public opinion is not new at all. Governments have paid attention to public opinion as long as there have been governments.
1. PUBLIC OPINION IN THE 21ST CENTURY DRAFT: NOT FOR QUOTATION OR CITATION Russell G. Brooker, Alverno College Todd Schaefer, Central Washington University “Methods of Measuring Public Opinion” “Truth reveals itself in degrees, and we can progress from an incomplete to a more and ever more complete comprehension of truth. Truth is not a thing, not an object that we either have in entirety or have not at all.” - Goethe The formal academic study of public opinion is relatively new but the practical study of public opinion is not new at all. Governments have paid attention to public opinion as long as there have been governments. Even the most oppressive tyrants need to know what the people are thinking, even if just to oppress them more effectively. As the famous investigator of public opinion, V.O. Key, Jr. said, “Governments must concern themselves with the opinions of their citizens, if only to provide a basis for repression of disaffection.” (1961, 3) One of the earliest expressions of public opinion was rebellion. Peasant rebellions have occurred throughout history. When the king saw his subjects in open rebellion, it was a pretty clear sign that the public’s support for his government was eroding. Unpaid taxes was another clue; when rulers saw their tax receipts dwindle and heard reports of tax collectors being killed, they knew that public opinion was turning against them. For centuries rulers have established secret police forces to find out which people oppose the government and to eliminate them. Secret police have, among other things, acted as monitors of public opinion. Fortunately, not all governments need to hire secret police or wait for peasant rebellions to learn about public opinion. Democratic governments, especially, have much better procedures to learn about public opinion and measure it. In this chapter, we explore ways that governments can learn about, and act on, their citizens’ opinions. We classify the methods of learning public opinion into informal and formal methods. The informal are very important but they do not involve any formal explicit research methodologies. Informal methodologies include elections, letters from constituents, media coverage, and protest marches. Formal methods, on the other hand, involve definite research designs and formal research methodologies; they are methods designed by experts to research public opinion. Examples of formal methodologies are telephone surveys, focus groups, and content analysis. In this book, like practically every book on public opinion, we devote most of the coverage to formal research methods, especially survey research. But because of our focus on how opinions held by members of the public get translated into public policy, we spend more time on the “informal” methods here than is customary in a book like this. We attempt to view public opinion as governmental decision makers would see it. In the real world of politics, these governmental leaders find out about people’s opinions in a variety of ways, both informal and formal. We will once again emphasize the informal ways that political leaders learn about public opinion in the chapter on political participation. We will see in that chapter that the informal methods tend to favor the wealthy and educated members of the society. 1
2. I. “Informal” Ways to Measure Public Opinion A. Elections The most common way for a democratic government to learn about public opinion is through elections. Elections are built into the system, at regular intervals in the United States and at irregular intervals in other democratic countries. They are important because they determine who staffs the government, and they are also one way for the public to express its feelings about politics. But they are not a particularly precise method for ascertaining public opinion. Politicians and researchers may try to figure out why one person won an election rather than another, but usually there are so many factors that it is impossible to single out one or a few and “the” reason or reasons. A candidate may have won because he or she is in touch with the voters and understands their needs better than any other candidate. But, on the other hand, a candidate may have won because he or she is a better speaker or because of more or better television advertisements. Even if a candidate won on “the issues,” it may be difficult to determine which issues were most important. Some voters may have decided on the basis of economic issues, while others decided on the basis of a candidate’s stand on abortion or gun control. In the end, all that elections tell us about “what the public wants” is that they preferred one candidate over another—not why. Elections are also imperfect measures of public opinion because they reflect only the opinions of those who voted. Certainly, in societies in which all adults have the right to vote, elections can reflect the various views of all the people. But in practice, not everyone votes—especially in United States, where only about half the eligible voters participate in presidential elections, and even fewer do so in other elections. Therefore, elections tend to reflect the views of those who vote, who are not necessarily representative of the public. As we shall see later, elections tend to reflect the viewpoints of more educated, wealthier citizens. While elections are a blunt measure of public opinion, they are the most effective means by which public opinion can control the government and public policy. Elected politicians, who are future candidates, will avoid taking especially unpopular positions on issues because they know that voters might notice and could show their displeasure at the polls. If voters do not notice a candidate’s position on their own, opposing candidates will be happy to point it out. In this way, elections serve to hold politicians accountable to voters’ opinions—even latent opinions that the voters may not consciously hold. Politicians’ perceptions of public opinion, then, may be as important as the actual public opinion itself. A state legislator may refrain from voting for an obscure but unpopular bill because she is afraid that constituents would be offended and vote for her opponent in the next election—even if, in fact, they would never know. B. Interest Groups and Lobbying: It may seem unlikely that interest groups would be valid measures of public opinion. They are remarkably unrepresentative of the public as a whole. The wealthy and the educated members of society are much more likely to be organized into interest groups and employ representatives. The poor and uneducated are much less able to speak to the government through lobbyists. Nevertheless, legislators, staffers, and other government personnel do pay attention to what interest groups say. They have good reason to do so. Good lobbyists tend to be well informed about their issues concerning their employers, they have access to facts necessary to write laws, they understand the political process, and they are present when necessary to answer questions. In a study of committee staffers in the Illinois state legislature, Susan Herbst (1998) found that the staffers did pay a great deal of attention to lobbyists and, in fact, considered what they said to be public opinion. Herbst was very clear that the staffers did not merely consider their views to be surrogates for public 2
3. opinion; they considered them to be public opinion. The staffers relied on lobbyists to be their conduit to understanding public opinion partly because they perceived them to be in touch with the people of Illinois and to honestly relay the people’s opinion. The staffers felt that the lobbyists were honest because their credibility would be ruined if they were caught lying. Another reason the staffers relied on the lobbyists to understand public opinion is that the lobbyists were always there and were knowledgeable about the issues and the bills being considered in the legislature. While the public might be “for” lowering a tax, ordinary people would not understand how a tax could be lowered or how much it could be lowered. The lobbyists could. The staffers, according to Herbst, saw ordinary citizens as lazy and uninterested in C. The Media: Many government officials, and many regular citizens, look to the media to understand the views of the public. In Herbst’s study, she found that legislative staffers also considered the media to accurately present public opinion. The media are important in understanding people’s opinions, and we will devote an entire chapter to the media later in this book. Media, such as television, newspapers, and magazines are important because of the news they choose and how they portray the issues. In other terms, they are important in determining the political agenda (what people in the government are thinking about) and in framing the issues (how the issues are being The print media are also important as conduits of opinions from editorialists, columnists, and ordinary people who write letters to the editor. Most large newspapers print the opinions of their editors and run the articles of liberal and conservative columnists. Most also print letters to the editor that that allow ordinary people the power to express practically any sort of opinion. Some magazines, such as Time and Newsweek cover politics in an essentially nonpartisan way, but there are many more magazines that represent practically any political point of view in the American political spectrum. Conservative magazines such as National Review express views right of political center, while liberal magazines such as The New Republic tend to explain politics from a more left-leaning point of view. These ideological magazines are useful for explaining their perspectives to their readers and to government officials. Conservative, or liberal, elected officials often look to conservative, or liberal, magazines for support and justification of their opinions. D. Letters and Calls: People use letters and telephone calls to express their opinions to their elected representatives. While many of these letters and calls are about specific personal problems, such as lost Social Security checks, many of them are about contentious political issues. Politicians notice when their constituents write. A few letters from constituents may represent the opinions of thousands of other voters. Letters and calls often spontaneously arise from interested ordinary people. But sometimes interest groups organize their members to write or call their representatives in state legislatures or in Congress. The interest groups understand how elected officials pay attention to contact from their constituents and use those constituents to “lobby” lawmakers. This is known as “grassroots” lobbying. In two examples from President Clinton’s administration: 1) interest groups in the student loan industry (the people who may have handled your loan) urged their members to contact their representatives and senators to fight against a bill that would reform the student loan procedures (Waldman); 2) interest groups in the health care industry, such as HMOs and hospitals, rallied their members to contact their representatives and senators to oppose Clinton’s national health insurance plan (Jacobs and Shapiro). While “grassroots” lobbying is not always successful, it was effective in these two examples; neither the student loan industry nor the health care industry was re-organized by Congress. 3
4. Again, letters are not a “fair” way to assess public opinion. Letter writer tend to be better educated and wealthier than average citizens. So the voices lawmakers hear in the calls and letters the receive from constituents tend to speak with higher class accents than most citizens. E. Protests: In democratic governments as well as dictatorships, protests have served governments as indicators of citizens’ dissatisfaction with government policies. In the United States, especially in the last 40 years, protests have been a staple way of communicating with the government. In the 1960s, civil rights leaders pioneered very effective use of protests, matching them extremely well with the emerging medium of television. The goal of protest is to get the media, and by extension the public and government officials, to notice a problem they have been ignoring. Martin Luther King, Jr. was a master a handling the media to magnify the impact of protests. King wanted his protest marches in the south to show the brutality of the Jim Crow, segregationist, political and legal system. He aimed his protests at the most visibly brutal Southern lawmen, such as Sheriff Jim Clark in Selma and Bull Connor in Birmingham, Alabama. By being brutalized on television, King and the other civil rights workers showed that the southern legal system needed physical coercion to main white supremacy. These protest marches were very successful; they turned the attention of the media—and Presidents Kennedy and Johnson and the Congress—to the southern oppression of African Americans and resulted in the passage of the Civil Rights Act of 1964 and the Voting Rights Act of 1965. Since the success of the civil rights protests, many other groups have adopted protests as part of their repertoire in communicating with the government. The women’s movement, the pro-life and pro-choice forces, and the gay rights movement have used protests over the last thirty or so years. But what was novel in the early sixties is commonplace today, and it is increasingly difficult to attract the media’s attention and to be noticed in protests. In the 2003 Iraq war, there were numerous anti-war protest marches, but they attracted minimal media attention and did not appear to change government policy at Again, protests are not especially useful for determining what the public as a whole thinks. While protest seems to be a method that anybody, even the poorest people, could use, in fact, those citizens who are better educated and have higher incomes tend to protest more than others. Protest, like the other “informal” methods of ascertaining public opinion, is skewed toward the well-off segments of the public. F. Straw Polls: Straw polls are a compromise between formal and informal methodologies. Straw polls typically look like formal public opinion polls, but they are conducted with minimal concern for the validity of the results. Before the advent of statistically-based surveying in the 1930s, all polling was straw polling. Today, in a common type of straw polling, members of Congress mail questionnaires to their constituents to ascertain their opinions on the important issues of the day. Or they tally the letters and telephone calls—so many people on one side, and so many people on the other side. Since there is no effort to make the people who answer these surveys—or call or write—representative of all the constituents, the results of the surveys and tallies typically do not represent the views of all the constituents. Nevertheless, research has shown that members of Congress often do use the results of these straw polls (Jacobs and Shapiro, 2000). In one imaginative use of straw polling in the 2003 governor’s race in California, Taco Bell announced that anybody ordering a beef burrito was casting a “vote” for Arnold Schwartzenegger, the Republican challenger, and those ordering chicken burritos were casting “votes” for Gray Davis, the Democratic incumbent. 4
5. II. “Formal” Ways to Measure Public Opinion “Formal” methodologies for measuring public opinion are usually classified into quantitative and qualitative approaches. Generally, formal methods are more systematic ways of ascertaining public opinion. Formal methods are likely to be conducted by scholars who understand their proper—and improper—uses and who are less likely than politicians or other non-scholars to misuse them or misinterpret them as standing for mass public opinion when they do not. Quantitative methods involve numbers—and usually statistics. Most public research is conducted quantitatively, almost always by surveys. However, much research on public opinion is conducted qualitatively. Although qualitative research methodologies tend to be less important in academic research, they are often extremely important when politicians and candidates conduct research for their own purposes. One qualitative methodology, focus group discussions, is used extensively by gubernatorial, Congressional, and presidential candidates as they develop and conduct their campaigns. A. Formal Quantitative Methods Let us begin this section of the chapter with a story from an early attempt to research public opinion through survey research. It is the story of the attempt of a magazine, The Literary Digest, to predict the outcome of the 1936 presidential election. It is a perfect bad example. That is, it demonstrates how bad methodology leads to bad, inaccurate, results. The Literary Digest was a popular magazine that had correctly predicted several elections before 1936. The magazine’s method of ascertaining voters’ intentions was to mail out “ballots” to millions of people and then count the “ballots” that were mailed back. In 1936, the magazine mailed about 10 million “ballots” to people who had telephones and owned automobiles. About two million people mailed them back. With great confidence, The Literary Digest announced that Alfred Landon would be the next Have you ever heard of President Landon? We think not. Mr. Landon received 36.5% of the popular votes and only eight electoral votes (Maine and Vermont). Keep this story in mind. We will return to it later as we discuss mistakes to avoid when conducting survey research 1. Sample Surveys The most common method, by far, for learning about public opinion is the sample survey. In a sample survey, researchers ask a few hundred or a few thousand people their opinions about the issues being considered. When applied to political use, such as in election campaigns, survey research is often called “polling” and survey research studies are called “polls.” Some researchers think of survey research as the only way to learn about public opinion, and they devote all, or almost all, of their analysis of public opinion to the analysis of survey research. We think survey research is the most important way to learn about public opinion, and we devote most this book to survey research and what analysts have found through survey research. In later chapters, we place great emphasis on two major sample surveys—the American National Election (NES) study and the General Social Survey (GSS). a. Types of surveys There are basically three ways to survey people: face-to-face interviews, telephone interviews, and mail 1) Face-to-face interviews 5
6. In face-to-face interviews, interviewers talk with respondents in person, usually in their homes. At one time, this was the most common type of surveying, but it is seldom used today. Two major exceptions, where face-to-face interviewing is still used, are the National Election Studies and the General Social Survey. The advantage of face-to-face surveys is that they can be very long (sometimes over an hour) and more complex, because the interviewer can explain the questions to the respondents. In addition, the interviewer can use visual aids, such as pictures or scales. Scales are often used in many types of surveys, but they are easier to visualize in a survey where the respondent can actually see the scale. For example, in one type of question the respondent places herself on a “thermometer” scale in evaluating somebody, such as an elected official. If the respondent really likes the official, she would place herself at “100,” but if she really disliked the official, she would place herself at zero. While this respondent could place herself on the scale in any form of survey, it would be easier in a survey where she could see a picture of a thermometer. There are two serious drawbacks to face-to-face surveys. The first one is that the respondents are unlikely to give embarrassing or socially unacceptable answers because they do not want the interviewer to think ill of them. Questions on race are especially sensitive in face-to-face interviews. Another problem is that they take a long time to complete. It could take over a month to complete a simple survey of 400 respondents. But the most important problem with face-to-face surveys is their cost. Considering that interviewers have to drive to respondents’ homes, and if they are not there have to drive back again, the cost of interviewers’ salaries can skyrocket. In the National Election Studies, the cost is several hundred dollars per completed interview. Largely to save money, the NES has included telephone interviews in the last few surveys. 2) Telephone interviews The most common type of public opinion survey conducted today is a telephone survey. If you read about a poll in a newspaper or magazine or hear about it on television, it was probably conducted over the telephone. Telephone surveys have some advantages over face-to-face surveys, especially in their much lower cost and faster implementation. A 15-minute survey of 400 respondents might cost $15,000 and take a few days. If speed were very important, it could be completed overnight. The most important drawbacks to telephone surveys are their simplicity and short length. It is impossible to ask, and receive meaningful answers for, complex or long questions. People simply have a difficult time processing complex questions over the telephone. The interviewers can help, but they cannot show pictures over the phone. Another drawback is the short length. Whereas face-to-face interviews can last over an hour, telephone interviews seldom take more than 20 minutes. (Some telephone surveys do take 40-50 minutes, but it is unclear how alert the respondents are by the end of the interviews.) Another problem is the unexpected intrusive nature of phone surveys; while it not true that every telephone survey takes place just as supper is beginning, it seems that way. 3) Mail surveys Mail surveys are seldom used in political research. Although they can be less expensive than telephone surveys, their drawbacks tend to be too important to overlook. The first main drawback is that the response rate tends to be very low, often under 30%, so that it is questionable how well the respondents represent the larger population. The second main drawback is that it is impossible to determine who actually answered the questions. Did the intended respondent answer the questions, or did his teenage daughter do it? Or did he get his buddies to help him and give consensus answers. It is impossible to tell. Two advantages of mail surveys are that: 1) pictures can be included to illustrate the questions; and 2) respondents can give embarrassing, socially unacceptable, but honest, answers without fear of being identified. For example, in a recent survey conducted for a state agency, people who had lost their driver’s licenses, mostly for drunk driving, were asked how often they drove illegally. Since the 6
7. questionnaires were not identified, the respondents knew they could answer the question without fear of repercussion. Most of them did report driving illegally, and 17% said they drove illegally every day (Brooker, 2003). This survey had the typical disadvantages of mail surveys in that the response rate was under 30%, and the researchers did not know for sure who answered the questions. However, the researchers were convinced that drivers without valid driver’s licenses would not answer honestly on the telephone and thought that a mail survey would yield the most honest answers.1 Of course, if the questionnaires are identified, as with ID numbers, this advantage does not exist. 4) Other methods of data collection Two other survey methods may be used to gather data about public opinion. Although they are sometimes used with business marketing research, they are rarely used in political surveys. The first is Internet interviewing. The questions are much like those in mail surveys, but they are answered on the Internet. Only people who complete questionnaires over the Internet, a small percentage of the American population, are eligible to participate in these surveys, so the samples are not representative of the entire United States. Several companies conduct Internet surveys, and they do make sure that the sample demographics are generally representative of Americans, but only the most computer-wise people participate. Internet surveys may some day be widely used in political research, but not yet. Another survey collection method is intercept interviewing. Intercept interviews are usually conducted in shopping malls and involve interviews of people who are “intercepted” while shopping. The interviews are similar to face-to-face interviews, although they are shorter. Businesses use mall intercepts every day. You may have seen them; the “interceptors” are usually well-dressed women with clipboards. However, intercept interviews are seldom used in political research. The obvious disadvantage is that only the people who are already at the intercept location can be “intercepted.” In malls, only mall shoppers can be interviewed, and they do not represent the general American public. b. The population and the sample Regardless of the method used, there are two important concepts to understand in survey research. The first is “population.” A population is all the people about whom one is gathering information. If one were interested in a presidential election, the population would be all the voters in the election. If one were interested in the opinions of women in Georgia, the population would be all the women in Georgia. The second concept is “sample.” A sample is the all people that one actually asks for their opinions. Typically, the sample is much smaller than the population. The reason that a sample is used is that it is almost always impossible or impractical to interview the entire population. The cost and logistics would also be prohibitive. Even in a town where only 5,000 people voted, it would not be possible to interview all the voters. In surveying, a small number of people, usually between 400 and 2,000, are asked for their opinions on issues. Researchers do not particularly care about the opinions of these specific “select few” or feel that their opinions are more important than those of others. Rather, researchers care about their opinions because they represent what the larger population thinks. In statistical language, the opinions of these few hundred or so people can be generalized or “projected” to the entire population. Surprisingly, it takes relatively few people to get an accurate picture of the whole; the best national surveys, such as the NES and GSS consist of interviews of only about 1,500 to 2,500 individuals. 1) Randomness The reason the researchers thought that people without driver’s licenses would not answer honestly is that they conduced some face-to-face preliminary interviews with drivers who had lost their licenses. These drivers told them that they would not answer honestly on the telephone but would in a mail survey if it were anonymous. 7
8. However, to obtain accurate results, not just any 1,500 people will do; the people must be selected randomly. The word “random” is very important in survey research. When a sample is selected randomly, every person in the population has an equal chance of being selected. The reason that randomness is so important is that a random selection is likely to be representative of the population from which it was selected, and its likely deviation from the characteristics of the actual population can be mathematically calculated. If a sample of people is selected randomly and then asked questions, we can be reasonably sure that the opinions expressed are going to be close to those of the entire population. There are two basic ways to select a sample randomly. One is through “simple random sampling” (SRS). In SRS, a researcher has a list of people who could be questioned and assigns a random number to each one. If the researcher needs 1,000 people in the sample, he or she selects 1,000 random numbers and surveys the people who are attached to those numbers. The other principal method of random selection is “interval sampling.” Using this method, a researcher has a list of people and selects every nth person. For example, if you wanted to sample 500 students in your institution, and there were 5,000 students from which to choose, with interval sampling, you would select every 10th person. In almost all surveys involving public opinion, the researcher does not have a list of people who could be included in a sample. For example, there is no list of potential voters to assign random numbers or to choose “every 3,000th one.” When there is no list, researchers can still select random samples. In these cases, sampling can be accomplished using multi-stage selection. In an in-person survey, such as the American National Election studies (NES) or the General Social Survey (GSS), the United States is divided into several areas, called Primary Sampling Units (PSUs). Some of those are chosen, with the likelihood of selection being proportionate to the population in each one (if there are 2 million in one and 4 million in a second, the second one is twice as likely to be selected). Each PSU that is selected is then divided into smaller regions, with some of those selected in the same manner. The smaller areas are then again sub-divided into city blocks or rural areas are chosen. Then people to interview are selected from those final blocks or areas. For telephone surveys, the country is divided into area codes and exchanges with some chosen randomly. Within each area code/exchange permutation (such as 555-382-….) some “hundred series” are selected. There are companies that specialize in telephone samples and know which “hundred series” are most likely to have residential telephone numbers, as opposed to business or government numbers). The best “hundred series” are selected and random two-digit numbers are added to them. For example, if 555-382 were the area code/exchange combination and “62” were the “hundred series” chosen, the random last two digits might be “43.” If that happened, the final number chosen for the sample would be 555-382- 6243. The number itself, not the people with the number would be selected. The interviewer would then have to choose one of the adults in the household randomly to be interviewed. If the number were not in service or were a business number, another number, chosen the same way, would be substituted. 2) How samples can be biased A sample represents the larger population only if each person in the population has an equal chance of being included in the sample. If everybody in the population is not given an equal chance of being selected in the sample, we say the sample is “biased.” The most famous example of a biased sample is the Literary Digest sample of 1936—the perfect bad example that we mentioned earlier in this chapter. Since the sample was taken from households that had automobiles or telephones, those without them had zero chance of being selected in the sample. Since 1936 was during the Great Depression, and only wealthier households had automobiles and telephones, the sample was biased in favor of wealthier voters, who were more likely to vote Republican. 8
9. Another way that the Literary Digest survey was biased was that the survey respondents themselves decided whether they would send back the postcards. Those most interested in the election were then the most likely to respond. This self-selection bias was a problem because those who choose to participate are different in some way than those who choose not to participate. There are several other ways a sample could become biased. If you, for example, wanted to survey a random sample of the students at your college or university but selected your sample from a list of students that excluded transfer students, you would have a biased sample because transfer students would have no way of being included in the sample. If you used a list of students that was a year old, your sample would be biased because it would not include freshmen or students who had transferred into your school in the last year (and it would include seniors who had graduated in the last year, whom you would not want). In the case of the multi-stage sample selection methods, a sample would be biased if some cities or areas of the country were left out of the selection process. The sample would also be biased if the populations were estimated incorrectly, so if an area of 20 million people were incorrectly listed as having a population of 2 million people, its chance of being included in the sample would be only one-tenth as large as it should be. In the case of a telephone survey, if some area codes or exchanges were left out, the resulting sample would be biased. Sometimes samples are biased through no fault of the researcher. For example, there could be a survey of residents of a state in which 20% of the adults had college degrees, but 30% of the sample had college degrees. In fact, it is not uncommon for more educated people to be over-represented in surveys because them tend to be more likely to cooperate with the interviewers. If there were a sample in which the college-educated population were over-represented by a 3-2 margin, a researcher could “weight” the sample by counting each college-educated respondent as 2/3 of a respondent. Therefore, if there were a sample of 500 respondents, and the college-educated were supposed to be 100 of those 500 (or 20%), but there really were 150 college-educated respondents, by counting each of them as 2/3 of a respondent, they would count as 100 respondents. If one weights the college educated, one must also weight the non- college educated. If they are under-represented so that there are 350 of them when there should have been 400, they get a weight of 400/350, so that each one counts as 1.143 respondents. In the end, there would be 500 weighted respondents, 100 of them college educated, and 400 not. Sometimes researchers intentionally create samples that are disproportionate. A researcher might do this to study a minority group in a population. For example, if a researcher wanted to compare the opinions of African Americans and whites in a particular state and blacks made up 10% of the state’s population, it would not make sense to interview 1,000 respondents to get a sample of only 100 blacks; comparisons of blacks and whites would be very difficult. It would make more sense to interview 500 black respondents and 500 white respondents. But if the researcher wanted to combine the 500 black and 500 white respondents to study the population as a whole, the combined sample would be biased in favor of blacks. The researcher could then weight the data so that black respondents would be given weights of 100/500, or .20, and whites would be given weights of 900/500, or 1.8. Most statistical software programs, including SPSS, have easy-to-use commands to weight data. In several of the following chapters of this book, we shall examine NES and GSS data; the respondents in these surveys have been weighted to take into account and correct biases in the samples. c. The questions After researchers choose the sample of people, they must ask the questions. Some questions, such as age or education, are easy to ask, and interviewers can use standard question formats. For example, the 9
10. question on party identification has been standardized to be asked in this manner: “Do you normally consider yourself to be a Republican, a Democrat, an Independent or what?” 1) How questions can be biased or misleading: Though some standard question formats exist, avoiding bias when asking questions on opinions is not so easy. There are many ways to write a biased question. A question may be biased because it portrays one side of an issue more favorably. For example, one Republican Congressman sent a questionnaire to his constituents that included the question, “Should the phased-in tax reductions passed in 2001 be sped up in an effort to stimulate the economy?” The option of speeding up the tax reductions is shown more positively than the option of not speeding them up—there is a reason given for speeding them up but not for not doing so. Questions can be even more biased if they use emotional words. For example, the question, “Do you favor killing unborn babies?” is unlikely to uncover anybody’s pro-choice opinions. Another way a question can be biased is to link one of the possible answers to an authority figure. A question from the same Congressman’s questionnaire asked, “Do you support the President’s efforts to completely disarm Iraq?” (Congressman Jim Sensenbrenner, News from Congress, April 2003) This question is more likely to find out whether the survey respondents like President Bush than if they think his policy toward Iraq is good. Sometimes if a question is difficult to understand, survey results may not be accurate. For example, in 1992, people were polled to see if they believed that the Nazi Holocaust had really happened. There was concern that some individuals who denied the existence of the Holocaust were convincing the public that Hitler had not tried to exterminate the Jews. The question to ascertain whether people believed the Holocaust had happened was, “Does it seem possible or does it seem impossible to you that the Nazi extermination of the Jews never happened?” People were concerned that 22% of the survey respondents said it was possible that the Holocaust never happened, and 12% did not know. Of course, the real problem was the question. A respondent had to say “impossible” to mean that it had happened. The question was asked later with a changed wording, “Does it seem possible to you that the Nazi extermination of the Jews never happened, or do you feel certain that it happened?” With this wording, 91% said the Holocaust had happened. (get citation) Sometimes a question may be worded well, but if the respondents are not competent to answer the question, the results will be meaningless. Take the following question: “With the latest changes in the Chinese government, people wonder who has the most power. Which of these men would you say has the most power, Jiang Zemin, Hu Jintao, or Li Ruihuan?” This is a reasonable question, but not for the American public because very few Americans have any idea who these men are.2 Sometimes, it is not clear how a question should be worded. In 1987 a Gallup poll was conducted with 4,244 American adults to get their opinions about American politics. One of the questions inquired about each respondent’s concern for the outcome of the 1988 election. The question was asked in two different ways, each to half the respondents. The question wordings and results are below: It is a good question, but we do not know the answer. Officially, Hu Jintao is the most powerful man in China, but Jiang Zemin has a great deal of informal power. Li Ruihuan is officially the second most powerful man. 10
11. Alternate Question Wordings Version 1: “Generally speaking, would you say that you personally care a good deal which party wins the presidential election in 1988 or that you don’t care very much who wins?” Care a good deal 54% Don’t care very much 40% Don’t know 6% Version 2: “Generally speaking, would you say that you personally care a good deal who wins the presidential election in 1988 or that you don’t care very much which party wins?” Care a good deal 76% Don’t care very much 20% Don’t know 4% Source: Orenstein et al. (1988) So, after reading the results of these two questions, how many Americans cared a good deal about the outcome of the 1988 election? The answer is 54% or 76% or somewhere in between. We don’t know any closer than that. Both these question wordings are “correct,” but they give us very different answers. We do not have a solution to this problem. Welcome to the world of survey research! Truth reveals itself in degrees. d. Sampling error and other errors The first thing to understand in sampling error is that there is no error—that is, there is no mistake. Sampling error is simply the statistical variation that sample survey results will yield. The idea behind sampling error is that if, say, we took a sample of 400 Americans and found that 54.5% of them approve of the job the President is doing, we would not insist that exactly 54.5% of all Americans approve of the President. We would realize that the true percent for the entire population might be 54.7% or 52.1% or 55.9%. All those percents are reasonably close to the 54.5% that we received. If we have conducted the survey correctly, we do not expect to get exactly the correct percentage, but we expect to be close. Sampling error just tells us how close we would be to the actual percent in the real world, with a certain probability of accuracy. Sometimes, people refer to the sampling error as the “plus or minus percent.” Usually, analysts use the 95% level of confidence. They say that a percentage is accurate plus or minus x% 95% of the time. For example, if the survey found that 54.5% of the respondents approved of the way the President was handling his job, an analyst would say that the true percentage was 54.5%, plus or minus 5% at the 95% level of confidence. That means that there is a 95% probability that the true percentage is between 49.5% and 59.5%. If we wanted to be more precise, we could interview more people. If we interviewed 1,100 people, we would be +/- 3% instead of +/-5%. The following table shows the increasing precision that comes with each sample size. Basically, more interviews is better; the main disadvantage of more interviews is the added cost. 11
12. Calculating Sampling Error Sample Size +/- “Error” 50 14 100 10 200 7 300 6 400 5 500 5 600 4 700 4 800 3.5 900 3.5 1,000 3.1 1,100 3 … 2,400 2 … 9,600 1 You will notice that with very small sample sizes, the error range is very large. If 52% of respondents in a sample of 50 respondents say they are going to vote Republican, we are 95% sure that the actual percent in the population who plan to vote Republican is 52% +/-14%, or between 38% and 66%. Not exactly shocking news! We probably knew that before we conducted the survey. Small samples are not very You may also notice that with 400 respondents the error margin is +/- 5 percentage points. That is why a sample of 400 is common for commercial surveys. Generally, academic surveys, such as the GSS and NES include around 1,500 – 2,500 respondents. This is the point where the added expense of interviewing more respondents becomes too large. The added expense grows larger as we add respondents. Notice how easy it is to reduce the error margin at first but how it becomes more difficult. In order to reduce the error margin from 14 percentage points to 5 percentage points, we need to add only 350 more respondents. But to decrease it from 5 percentage points to 1 percentage point, we must add 9,200 respondents. But sampling error is not the only source of error in surveys. The 1936 Literary Digest survey mentioned previously illustrated some common errors. First, the Literary Digest committed a population specification error. That is, the magazine specified the population of voters as those households that had cars and telephones. In fact, many people, especially the poor and lower middle class, did not have either cars or telephones. Second, the magazine committed selection error because the actual respondents were not selected randomly; instead, those most interested sent back their postcards. In any telephone survey, the people who do not have telephones are excluded, so telephone surveys have inherent selection error. However, since over 90% of Americans do have telephones, and other types of surveying are too expensive, researchers have learned to live with the exclusion of non-telephone households. Another way of explaining the problem of selection error is non-response error. This simply means that the people who did not respond are different from those who do. If the people who did not respond were exactly the same as those who do, non-response would not matter because the respondents would be a good representation of all the people in the population. 12
13. Two other errors that did not happen in the Literary Digest survey, but could happen, are interviewer bias and processing errors. Interviewer error occurs when interviewers do not ask questions properly or suggest answers improperly. For example, an interviewer may not like the current president and may cause the respondent to rate the President’s behavior in office unsatisfactorily. Processing errors occur when answers are simply recorded improperly; an example would be if a respondent evaluated the president’s behavior as “good” but the interviewer recorded it as “excellent.” In the Literary Digest survey, there were no interviewers, and while there may have been processing errors, they would not account for the wild inaccuracy of the survey’s prediction. Another possible error is instrument error. This type of error occurs when a question is biased or is not asked correctly. If the question leads respondents to answer one way, it is biased and does not give the researcher accurate results. We have listed some examples previously. To some extent, some instrument error is unavoidable when asking about something other than demographics and very clearly understood issues like voting. There is no one right way to ask for someone’s opinion on race, abortion, taxes, or any other important issues. There are several ways, all good, to ask about important issues. As we have seen, very small changes in question wording can yield vastly different question responses. e. Advantages and Disadvantages of Survey Research 1) Advantages Survey research is the method used most often to learn about people’s political opinions. It is also the research method we will use most often in this book. The clearest advantage to this method is that, with careful random selection of the sample, the results of survey research can be projected to the entire population. If 55% of 400 survey respondents say that they approve of a program to improve public education, a researcher can be reasonably sure that about 50%-60% of all the adults in the US population approve of the program. The main strength of survey research is that it enables a researcher to “project” or “generalize” to the entire population. There is simply no other way to learn what an entire large population, such as the American electorate, thinks. When used properly, surveys can describe the population or predict behavior very accurately. Much of the work of the U.S. Census Bureau involves sampling people, and the numbers generated from those samples have great credibility. The unemployment rates and inflation rates, for example, are derived from sample surveys. In elections, there are many polls that predict very accurately how the vote will go. There are some horror stories of wildly bad predictions, but reputable pollsters regularly predict election results within three or four percentage points. Another important advantage of survey research is that it is fast and can be relatively inexpensive. For less than $20,000, one can find out approximately what the American public thinks about an issue—and one can do it in a few days. 2) Disadvantages Survey research has its disadvantages and its critics. Taeku Lee, in Mobilizing Public Opinion, criticizes what he calls “the sovereign status of survey data” (2002) He says that the findings of polls, an “indicator” of public opinion, become conflated with public opinion itself. He quotes Susan Herbst: “These days we tend to believe that public opinion is the aggregation of individual opinions as measured by the sample survey…this definition is now hegemonic; when most of us consider the meaning of public opinion, we can’t help but think about polls or surveys (2002, 73) (also quoted from Herbst, 1995, 90). 13
14. The most common criticisms of survey research concern its superficiality, the lack of time respondents are given to reflect on the questions, the lack of information respondents have in answering the questions, the individualization of responses, and problems in operationalizing concepts. a) Superficiality: A major problem with survey research is that it is superficial. While a researcher can learn approximately how someone feels about an issue, the researcher cannot learn the complexity or subtlety of that person’s opinion. For example, if a survey respondent rates the president’s behavior as “good” rather than “excellent,” “fair,” or “poor,” it is not clear why the person gave the president that rating. Maybe the person thinks the president is “excellent” on the abortion issue, “poor” on taxes, “good” on foreign relations—or maybe the person thinks the president is “good” compared to the person’s expectation that he would be terrible. While it is possible to ask about each issue area separately, researchers seldom do more than ask about overall “domestic” or “foreign” issues. Even if a questionnaire asked a person for an evaluation of the president’s program on, say, education, it could never get at the complexity of the person’s thought processes. b) Lack of time to reflect: Another problem with survey research is that interviewers record what the respondents says on the telephone with only a few seconds to respond to each question. And the interview is almost always a complete surprise to the respondents, so they have no time to prepare. (It is extremely unusual to warn respondents ahead of time that they are going to be surveyed, partly because the researcher usually does not know who will be surveyed ahead of time.) c) Lack of information by respondents: Survey respondents frequently do not know the background or context of many opinion questions. Interviewers seldom give more than one or two sentences, if that, of background information. The researcher cannot know the context that respondents use in answering the questions? For example, does each respond understand the alternatives? Does the respondent know the implications of each response? When being asked about a proposed program, does the respondent know how much each program component will cost? Would the respondent answer the same if she knew how much her taxes would increase if her “favorite” alternative were actually enacted into law? Robert Weissberg (2002) conducted a survey concerning proposed new programs and found that fewer respondents favored programs when they were told how much they would cost. Even when issues are well known, well understood, and discussed widely, it is difficult to ascertain what people think about them. Take the issue of affirmative action at The University of Michigan as an example. This issue was widely known discussed in 2003, especially since President Bush spoke out against the university’s practice of affirmative action. But how does one measure people’s opinions on this one specific case? Does one measure opinions on affirmative action for members of racial minorities? What about affirmative action for students from poor economic backgrounds, for scholarship athletes, or for men planning to major in nursing?3 Should one measure the difference in opinions about affirmative action when used as quotas compared to when quotas are not used? Should one differentiate between affirmative action programs that use “bonus points,” as one of Michigan’s programs did, from those that do not use “bonus points,” as another Michigan program did? How many bonus points would be appropriate; would 10 points out of 150 be okay, or would 20 be acceptable? Should “bonus points” be cumulative? If an applicant is a poor black scholarship athlete, should he or she be entitled to three affirmative action considerations? These are all legitimate questions one could ask when attempting to measure the public’s opinions about one specific issue at one specific university. We have not addressed affirmative action in such areas as business or for other purposes, such as salaries or promotion. The entire issue of affirmative action is The University of Michigan offers “bonus” points in admissions—affirmative action—for these applicants, in addition to several other types of students. 14
15. much more complex than we have described here. Many issues, and practically all interesting issues, are very complex. It is not at all clear how one would go about measuring public opinion on such issues. d) Individualization of responses: In all sample surveys, interviewers talk with each respondent individually, one at a time. This is good in one sense, because no respondent can influence another respondent’s answer. But in the political world, people do influence each other’s opinions. We do not know is if the respondents would answer the same way if given some time to think or discuss the president’s program with friends. Critics of survey research have said that trying to measure public opinion by asking people for their opinions really does not measure public opinion as much as merely count “aggregations of atomized individuals.” (Herbst) They say that opinion is not really public unless people have opportunities to discuss their views with others. When people talk among themselves and read and hear others discuss their opinions, they are likely to become more informed. Robert Lane explained how people can refine their opinions by discussing them with friends and acquaintances at "The hollering, the bantering, the kidding that Republicans take from Democrats and vice versa— these experiences bring the ‘way out’ opinion nearer to the group norms, but while they homogenize opinions, they may also, in the American work culture, civilize them too.” (Lane, Political Ideology, p. 241) e) Problems in operationalizing concepts: Survey research only approximates—or infers—public opinion, and its accuracy is heavily dependent on the specific opinion it is measuring. For opinions that are understood well by the public and are easy to operationalize, survey research can be very accurate. The best example of accurate survey research is election polls; everybody knows what an election is and knows what it means to vote for a candidate. Another example is a demographic question. If a respondent says she is 23 years old, we do not need to infer what “23” means. For information like this, surveys are excellent sources of information. But when the questions are more abstract, the approximations become less precise—and the inferences become less direct and convincing. Here, we do not mean “inferential” in the sense of “inferential statistics” in which one infers a population characteristic from a sample value. Instead, we mean that we infer the meaning of an answer in the “real world” from the exact answer that the respondent has given. Take, for example, a standard question in National Election Studies (NES) surveys about the government taking responsibility for providing jobs and a good standard of living for everybody. The question asks each respondent to take a position from “1” to “7” on this issue, with “1” meaning that the respondent thinks the government does have a responsibility, and “7” meaning that the government does not have a responsibility. This question has been used in the NES for many years. The precise wording in the 2000 NES was: “Some people feel the government in Washington should see to it that every person has a job and a good standard of living. Suppose those people are at one end of a scale, at point 1. Others think the government should just let each person get ahead on their own. Suppose these people are at the other end, at point 7. And, of course, some other people have opinions somewhere in between, at points 2, 3, 4, 5, or 6. Where would you place yourself on this scale, or haven’t you thought much about this? In the 2000 NES, 88 respondents of 1,807 asked the question placed themselves at point “2.” We think it is a safe assumption that none of those 88 people thought of themselves as having an opinion of “2” on this issue before the interviewer showed them the “1 to 7” scale. The “2” is just a convenient way to classify these 88 people as having liberal positions on this issue. We do not think that this is a bad way to categorize people’ opinions. In fact, we think it is a very good way; this question is very useful in classifying opinions of the American public, and we use this question several times in this book. 15
16. However, we are inferring the opinions of those 88 people; we do not know what they think about the responsibility of the government in Washington to see to it that every person has a job and a good standard of living. Most researchers who attempt to divine public opinion recognize the difficulties involved. They write questions that attempt to discover people’s opinions, aware that no question gets at everything and that no wording is perfect. The researchers know that no one question will get precisely the “true” opinion of a person. So they re-write questions to get at the same opinion in a slightly different way. They realize that then never get exactly the opinions of the people. One way around the problems with survey research is to complement it with other types of research. We will now discuss some other types of research. First we will look at other types of quantitative research and then at qualitative research. Box 2.1 What do survey answers mean? Goes here 2. Other Quantitative Methods Two other quantitative methods, experiments and analysis of mail, reveal different kinds of information than survey research does. These methods permit researchers to ask different kinds of questions than would be appropriate for surveys. The main problem with these methods is that, for the most part, they are not projectable, as surveys are. We cannot say, for example, if X% of subject (as experiment participants are called) in an experiment responded in one way to a case on gender discrimination then X%, or any other percent, of all adults would respond the same way in the same circumstances. The sample sizes are typically too small and study participants are typically not chosen randomly from the entire population being studied for the study results to be projectable. a. Experiments Experiments in public opinion research have yielded some very interesting and useful results. Most, but not all, experiments have been conducted in laboratories. In his book, Persuasion and Politics, Michael A. Milburn describes several laboratory experiments relevant to public opinion. In one, (pp. 78-79) he described a 1986 experiment by Lodge and Hamill in which subjects were classified into three groups based on their answers from an NES survey on interest and knowledge about political parties and elections. The experimenters then had the study participants read a fictitious biography of “Congressman Williams” along with some statements attributed to him. The statements were either consistent or inconsistent with Williams’ party label. For example, a consistent statement had Williams as a Republican opposing government programs to create jobs while an inconsistent statement had him as a Republican favoring such programs. The subjects were then given an unrelated task to distract them from the experiment. Then Lodge and Hammil gave a test to the subject that measured their recall of the of the policy statements. Not surprisingly, the most interested and knowledgeable study participants (the top one-third) recalled more of the information on the test than the least sophisticated participants (the bottom one-third) did. They had a base of knowledge and a conceptual framework to help them understand the information about “Congressman Williams.” However, the most top third was a little less accurate in recalling inconsistent information. They actually recalled more consistent information than was presented. Their conceptual framework of Democrat = Liberal and Republican = Conservative caused them to “correct” inconsistent issue positions. Milburn also pointed out a study on feminism and ideological consistency (pp. 103-04). Sherman and Gorkin (1980) first questioned subjects to determine their feminist attitudes and then asked them to explain a puzzle: 16
17. A father and his son are out driving. They are involved in an accident. The father is killed, and the son is in critical condition. The son is rushed to the hospital and prepared for the operation. The doctor comes in, sees the patient, and exclaims, “I can’t operate, it’s my son!” After this puzzle, the subjects were then asked to respond to a hypothetical case about a woman who had been turned down for a university teaching position because of her gender. Sherman and Gorkin found that subjects with strong feminist attitudes who had failed to solve the puzzle4 increased their support for affirmative action. They concluded: “When an important part of one’s self-image is challenged by one’s own behavior, and that behavior can neither be denied nor misperceived, a desire to reestablish that important self-image lead to the adoption of an extreme behavior to counteract or invalidate the implication of the threatening behavior.” (p. 397 in original) Sometimes experiments are conducted in conjunction with survey research instead of in a laboratory. These studies, unlike laboratory experiments, are statistically projectable to the entire population if the survey respondents are chosen randomly. In a 1991 article, Paul Sniderman and others described a survey research experiment they conducted to test the thesis of “the new racism.” That thesis holds that white people disguise their prejudice by giving socially acceptable answers concerning their racial opinions toward African Americans. According to the thesis, whites “will not say they are opposed to blacks getting help from the government because they are black; they will instead say they are opposed because blacks are not making a genuine effort to solve their own problems—the kind of effort everyone should make.” To test that thesis, the authors surveyed 1,113 white adults in the San Francisco bay area. In the survey, respondents were asked how much help the government should give a laid-off worker.5 Respondents were asked about a person who had five characteristics. The laid-off worker was: • White or black • Male or female • In his/her early twenties, mid-thirties, or early forties • Single, a single parent, married, or married with children • A dependable worker or not a dependable worker These five attributes were matched randomly so that whites were no more likely than blacks to be male or female or dependable or undependable—or have any other attributes. As a result, there were as many white female young single dependable workers as any other possible combination. Survey respondents were not told that they were being questioned principally about race. If the “new racism” thesis were accurate, many whites would favor government help for white laid-off workers but not for black workers. The authors hypothesized that conservatives would be especially likely to exhibit the “new racism. After the survey was completed, however, the authors found no such thing. Whites overall were not more likely to deny government help to blacks than to whites; in fact, the conservatives were more likely to favor government help to blacks than to whites. Sniderman et al. concluded that the survey results did not support the “new racism” thesis. b. Analysis of Mail In Mobilizing Public Opinion, Taeku Lee makes a convincing argument that public opinion can be read by studying letters written to politicians. He studied letters that individuals had written to US presidents from 1948 to 1965 about racial issues. He shows how studying the letters reflects the change in public The doctor was the boy’s mother. You knew that. The study had several parts. Just one part is described here for illustration. 17
18. perceptions about race over those 17 years. According to Lee, the problem with using survey data to study momentous change is that questionnaire writers are often behind the times; they cannot, and should not be expected to, anticipate great changes in the political agenda. Letters to politicians, on the other hand, often anticipate agenda changes and can react very quickly to changing issues. Lee points out how letters changed quickly both in number and content in reaction to political events. He explains how the letters reflect the opinions and the writers’ frames of reference. His data show how often the most common themes occurred in the letters. The following table contains his data concerning letters written to presidents from 1960 through 1965 from African Americans, anti-movement white Southerners, and pro-integration (usually non-Southern) whites. The differences in their perspectives are clear. While the African American and racially liberal white letter writers emphasize universal rights more than any other subject, the Southern anti-movement whites mention communism, states’ rights and claim that the movement activists are communists. Most Common Individual Frames 1960 - 1965 Type of Correspondent Frames African Southern Racially Liberal Americans Whites Whites Universal rights 34% -- 36% Justice and equality 24 -- 13 Democratic principles 11 -- -- Black identity, interests 15 -- -- Religious morality 12 8% -- World opinion 12 -- 19 Political symbols 11 -- -- Communism, cold war -- 24 -- States’ rights -- 18 -- Civil rights activists as communists -- 17 -- Whites’, taxpayers’ rights -- 17 -- Movement leaders, organizations -- 13 -- Religious essentialism* -- 6 -- Police brutality -- -- 19 Equal protection under the law -- -- 15 Number of letters 459 892 *”Religious essentialism” refers to the claim that the Bible says that whites and blacks are essentially different. SOURCE: This table is abridged from three tables on pages 158,163, and 169 of Mobilizing Public Opinion He also claims that analysis of the letters gave more information than answers to survey questions. He writes, “These individual letters….revealed not only the views of correspondents about civil rights and racial integration but also the choices they made concerning the wording and framing of those views. Thee choices showed the powerful rhetorical force of dearly head values and the resounding influence of group-based interests, conflicts, and animosity.” Lee does not claim that letters accurately capture the opinions of the entire public. Letter writers are not representative of all adults; they tend to be better educated and more interested in the issues—especially the issue they write about—than most people are. However, Lee says that letter writers are a reasonable 18
19. measure of the opinions of the interested mobilized public, a layer of people between the elites and the mass public. III. Formal Qualitative Methods to Learn Public Opinion: Surveys, most experiments, and Lee’s mail analysis are quantitative; that is, they involve the use of quantitative, or numerical, data. But some research methods are qualitative; they involve the use of non- numerical data such as people’s ideas, observations, or opinions stated in their own words. The purpose of qualitative research is to uncover the quality of opinion. While quantitative analysis is very useful in determining how many people have a specific opinion, it is not very good, except superficially, in determining how people think and why they think the way they do. Qualitative research has the opposite strengths and weaknesses; it is useless in determining how many people have a specific opinion, but it is very useful for uncovering how people think and why they think the way they do. There are many types of qualitative research, but the most important ones for political research are in-depth interviews and focus groups. A. In-depth Interviews In in-depth interviews, researchers talk with people, usually face to face, for much longer times than they would in survey research—often several hours, which may be divided between days. The main purpose of in-depth interviews is to uncover the outlooks and perspectives of the people who are interviewed. Instead of asking a respondent to place herself on a seven-point scale on affirmative action, a respondent in an in-depth interview might be asked simply how she feels about affirmative action. She would then give her opinion and her perspective on the issue. Instead of labeling herself a “5” on a scale, she would explain what she thinks “affirmative action” means, when, if ever, it should be used, and to whom, if anybody, it should be applied. She might, for example, say that the government should develop affirmative action programs for poor people and for Vietnam War veterans but for nobody else. She might claim that numerical quotas are good, bad, or she may not even have an opinion on quotas. She would also explain the reasons for her opinions. A respondent in an in-depth interview might talk for 30 minutes on one subject if she really cares about it—or for less than a minute if she does not. The data obtained from in-depth interviews can be very nuanced and subtle. One can learn the context of respondents’ opinions and how they structure the world in their minds. A researcher can learn how somebody really thinks about affirmative action rather than just learning that she is a “5.” Of course, one cannot conduct enough in-depth interviews in one research project for statistical reliability; a typical number of respondents is 15. Therefore, the findings of in-depth research studies are not projectable to the entire population under investigation. In in-depth research, as in qualitative research in general, there is a trade-off of statistical reliability for more textured and contextual data. Jonathan Schell, a journalist, interviewed members of one family and their neighbors repeatedly over the course of the 1984 election. He summarized the advantages of his method in this way: “I knew that on the basis of my talks with these people I could not make political generalizations of the kind that political polls allow, but I wanted to find out things that a poll could not reveal. Instead of finding out a little about a lot of votes, I wanted to find out everything I could about a handful of votes. After all, I thought it was here—in the minds of individual voters—that an election, in the last analysis, took place.” (Schell, 4) 19
20. Probably the most famous study of this type is Political Ideology by Robert E. Lane. In this study, published in 1962, Lane talked with 15 “average”6 American men from one neighborhood in an Eastern seaboard American city (that he calls “Eastport). All the men were white married fathers. Lane talked with these 15 men over several evenings for several months in 1957 and 1958. In the end, he compiled approximately 3,750 pages of verbatim transcripts. Lane principally sought to uncover the men’s latent ideologies, the sources of those ideologies, and how they supported or weakened the institutions of With his extremely textured and detailed data, Lane examined the ideas and ideologies of the 15 men from several perspectives. To illustrate his approach, we can look briefly at his Chapter 9, on the views of the 15 men on the role of government, and their lack of cynical opinions. Lane found that the men felt that public officials served their own—the men’s—interests by serving the public welfare. Because the men believed that public officials tried to serve the interests of the people, they thought that the people are sovereign—and that individuals, including the men, are important. Lane concluded that the men believed, although they did not explicitly say so, that the people “run things by having their welfare serve as the criterion of policy.” (p. 160) The men felt that government officials cared about ordinary people, which is what mattered to the men themselves. Lane shows how qualitative research can be used in the political world in the advice he gives to conservative interest groups trying to influence the attitudes of “average” people: “The fact seems to be that the antigovernment, anti-Congress educational campaigns of the National Association of Manufacturers, the steel, oil, and utility interests, and the medical profession, have been employing the wrong theme. The men of Eastport may—and some do— think of congressmen as spendthrift, not too efficient, talkative, and sometimes obtuse—but they do not think of them as cold, unreceptive to communication, and indifferent to their fates. And these are what matter to them.” (pp. 158-59) In a 1980 book, We Shall Not Overcome: Populism and Southern Blue-Collar Workers, Robert Emil Botsch interviewed 15 textile workers in a Southern city. All the workers were aged between 18 and 33 years old and had similar blue collar jobs in two textile mills. Ten were white and five were black. Botsch questioned them about politics, with particular emphasis on economic and racial issues. His methodology was consciously modeled on Lane’s, even to the extent of interviewing 15 men. In a chapter on economic issues, Botsch shows how qualitative research can enhance, explain, and further develop quantitative findings. When asked, 14 of the 15 men said they favored to some extent the government providing jobs for people who want to work but were unable to find jobs. But, he noted, the men’s opinions on this issue were very complex. For example, one of the men first said “sure” when asked if the government should provide jobs for those who want to work. “But when asked whether he minds having his tax money spent for this sort of thing, he quickly began to back away.” (p.62) Botch pointed out that an analyst would have to examine the details of the men’s opinions to really understand their opinions on the government providing jobs. He then summarized their opinions in a more nuanced “A summary of their opinions on this issue would be close to the following proposition: The government should provide jobs if (1) merit is proved by the potential employee’s willingness to take jobs like those that these men have been willing to accept in their lives; (2) jobs such as these Fourteen of the 15 men had not attended college. The fifteenth had completed graduate training. Lane interviewed that man because he had been selected by Lane’s random selection method. However, he did not consider him “average’ enough and devoted little of his analysis to him. 20
21. are not available (most of the men believe they are available) and there is nowhere else to turn. A majority of these men, both black and white, also believe that government-provided jobs should not be of higher quality than those they endure. Enthusiasm is notably lacking for increasing taxes to finance the creation of such jobs.” (p.63) Box 2.2 on diversity of opinion goes about here B. Focus Groups Businesses use focus groups in marketing research every day. They are also used by political candidates on a regular basis. However, they are seldom used by academic researchers. Since the principal theme of this book is the translation of public opinion into public policy, it is important to include focus groups in this discussion because many politicians, largely in their roles as candidates, use them to learn about public opinion. A focus group involves a small group of people, generally about 8-12, and a moderator. The moderator asks questions, almost always open-ended. A typical question might be, “What do you think about the President’s policies?” The main purpose of focus group research, similar to in-depth interview research, is to find out what people think about important issues and how they think about them. That is, what perspectives do they use when thinking about issues? How do they frame the issues? What issues do they think are important? In his first campaign for governor of Wisconsin, Republican Tommy Thompson’s organization used focus groups to discover that many people in Milwaukee County were concerned about a new prison being built in the county. Thompson then ran against building a new prison in Milwaukee County. Focus groups helped him find this issue.(find the citation) Focus groups are also used to try out new issues stands or emphases. In the 1988 presidential campaign, George Bush’s organization conducted several focus groups in New Jersey to evaluate using Willie Horton as a campaign issue. Willie Horton was a convict who had escaped on a furlough from a prison in Massachusetts and subsequently raped a woman. The Democratic candidate was Michael Dukakis, the Governor of Massachusetts. The Bush focus group researchers found that if they brought up the subject of Horton and explained that Dukakis, as Governor, was responsible for lax incarceration policies in Massachusetts and therefore responsible for Horton’s crime, that several voters who had planned to vote for Dukakis changed their votes to Bush. (Scheiffer and Gates, 1989, pp. 360-63) Candidate Bush then attacked Dukakis as being responsible for Horton’s crime to great effect in the campaign. C. Advantages and Disadvantages of Qualitative Research 1. Advantages The principal advantage of qualitative research is that it can uncover the perspectives and thought patterns of ordinary people. Researchers can not only find out what people’s opinions are but also how they think of those issues and why they have the opinions they do. Whereas in quantitative research, the number of people with an opinion is important, the reasons for those opinions are secondary. For example, if a researcher finds that 36% of the respondents to a survey support a woman’s right to an abortion, that researcher is unlikely to find out why each of the respondents has the opinion he or she does. The proportion is the important thing. A survey researcher will probably not find that Respondent #135 supports the right to abortion because she sees it as a “right to privacy” issue, while Respondent #257 supports the right to abortion because he feels it is important to obey the current law, Respondent #445 actually opposes abortion but answered that she favors it because she sees the difficulty of trying to enforce an anti-abortion law, and Respondent #543 really had not thought about the issue before but says he favors it because he does not want to appear uninformed and has to say something. In most quantitative research, the numbers are compiled and whether respondents have good reasons for their opinions is not considered. 21
22. However, in qualitative research, study participants are asked to explain their opinions. In focus groups, participants can disagree with each other. The reasons for opinions are examined, and a researcher can gain a fuller understanding of the people’s viewpoints and opinions. In the case of abortion, the participants can explain their positions, and researchers can find out who supports abortion from a “right to privacy” perspective and who supports if from an enforcement or other perspective. Another advantage of qualitative research is that researchers can see what it takes for people to change their opinions, as they did in the Bush campaign. While it is possible to examine in a survey how people’s minds change as they encounter new evidence ( See Weissberg), it is hardly ever done. It is done much more easily, and often, in qualitative research. Qualitative research can answer questions that researchers never thought to ask. Discussions often stray into areas the researcher had not intended to cover. In a quantitative survey, it is practically impossible for a respondent to answer a question that is not asked. 2. Disadvantages The principal disadvantage of qualitative research is that results are not projectable. Qualitative results are not generalizable for two reasons. First, the subjects are seldom selected randomly; people who fit the requirements of the study are chosen without any effort to make sure that everybody in the population has an equal chance to be selected. Even if the subjects were selected randomly, the sample sizes are too small to use quantitatively (the 95% confidence range for a random sample of 15 respondents is +/- 26 percentage points, which is much too large to be useful). Because of the lack of generalizability in qualitative research, if six of ten participants in a focus group have an opinion, we cannot responsibly say that 60% of people in the population have that opinion. In fact, we cannot say any percent of people have that opinion. The correct percent may be 4% or 97%; we don’t know. We can find out what people think and why they think the way they do, but we cannot learn how many think that way from qualitative Box 2.3 on Hard Numbers goes about here. IV. Who Conducts Public Opinion Research? Three main types of organizations conduct public opinion research: • Academic research organizations • The media • Politicians • Commercial companies A. Academic Research Organizations Many colleges and universities have research centers. They conduct a wide variety of research, including political research. They may conduct research strictly for academic purposes, or they may conduct commercial research for paying customers. Generally, they are managed by professional researchers with student helpers who learn how to conduct research by doing it. Two of the most important academic research organizations are The Survey Research Center at the University of Michigan and NORC at the University of Chicago (NORC was formerly known as the National Opinion Research Center). The Survey Research Center is part of the Institute for Social Research. It has conducted interdisciplinary social and political research involving the collection or analysis of data from scientific sample surveys for more than 50 years. Its most important study for our 22
23. purposes has been the American National Election Study. This study was first conducted in 1948 and 1952. Since 1956 it has been conducted every two years at the time of Congressional and Presidential elections. The National Election Study is the premier source for public opinion over the last half century, especially as it relates to voting. The 2000 National Election Study is cited extensively in this book, and a subset of data from it is included on the books web site for your use. NORC has offices at the University of Chicago and in Washington, DC. NORC’s clients include government agencies, educational institutions, foundations, other nonprofit organizations, and private corporations. For our purposes, the most important study that NORC conducts is the General Social Survey. This survey has been conducted almost annually since 1972. It includes questions on a variety of topics; although it is not principally a political survey, it includes many political questions. Both the National Election Study and the General Social Survey use face-to-face interviews, although the National Election Study has recently used telephone interviews to supplement its sample. Both studies include questions that use the same wording year after year so that trends can be tracked. Although they phase questions in and out of the questionnaires to account for changing issues, they keep many questions the same so that people’s answers can be compared. B. The Media The media also conduct many political surveys. Almost all of the media surveys are conducted using telephone interviews. Media surveys typically are conducted to create articles for the media audiences. Larger media often have political polls during electoral campaigns and when major stories occur. Some major media collaborate on a regular basis; for example, the New York Times and CBS News frequently sponsor polls together, as do the Wall Street Journal and NBC News. Although the media may conduct their own interviews, they typically hire marketing research companies to conduct them. If you see a poll story in a newspaper, look for the “How Box.” This is a box that explains how the survey was completed; it usually contains the dates of the interviews, the name of the company that actually conducted the interviews, the sample size, and the margin of error at the 95% level of confidence. Box 2.4, the “How Box” goes here. In addition to conducting studies on contract, some commercial companies conduct regular surveys and sell the results to subscribers. The Gallup Organization (also known as the American Institute of Public Opinion) is the most famous of these companies. In fact, George Gallup was largely responsible for the rise of polling in the 1930s. In 1936, he challenged the accuracy of The Literary Digest and sold his polls to newspapers with a money-back guarantee that his results would be more accurate than The Literary Digest’s results. Of course, he was more accurate and did not have to refund any money. Although media and commercial polls are not conducted in order to further academic research, academic researchers have access to their results and use them in their analyses. In fact, we will cite some of these surveys in this book. Many research studies conducted for media are very good, but some are very bad. Sometimes television stations sponsor call-in surveys (“Push 1 for “yes” and 2 for “no.”) Sometimes they sponsor Internet surveys in which viewers are free to respond as many times as they wish. These surveys have been called SLOP, for Self-selected Listener Opinion Polls. Although the television stations usually warn the viewers that these are not “scientific” polls, they treat them as if they were. Be aware that these polls are not 23
24. scientific or accurate; in fact, they are worse than no information at all because they give the false impression of real information. You would do as well to make up your own numbers. C. Politicians Politicians and political parties conduct polls regularly, during campaigns and at other times. Candidates also sponsor focus groups. In fact, along with media polls, party- and politician-sponsored polls, focus groups are some of the principal ways that political decision makers learn about public opinion—and therefore are some of the main ways that office holders gather public opinion that is translated into public policy. However, research conducted by politicians is often undertaken in order to find ways to change public opinion rather than to learn it. Politicians use surveys and focus groups to find word and phrases to use in selling their preferred policies to voters (Jacobs and Shapiro, 2000). D. Commercial Companies There are many commercial research companies that work for media clients, political parties, and individual politicians. Some of these that conduct research for parties and politicians work for both Democrats and Republicans, while others work only for one party. These companies typically also conduct research for non-political clients, such as regular business firms. The following list shows commercial marketing research firms that conduct surveys or focus groups for both parties, for Republicans only, and for Democrats only. Each firm in the list is accompanied by a brief description taken from its web site; no guarantee is made here about the accuracy of the claims. 1. Companies that work for both Republicans and Democrats • Roper Center for public opinion research - Roper Center for public opinion research provides access to the world's largest archive of computerized social and political science data. http://www.ropercenter.uconn.edu/ • Public Agenda is a nonpartisan, nonprofit public opinion research and citizen education organization based in New York City. It was founded in 1975 by social scientist and author Daniel Yankelovich and former Secretary of State Cyrus Vance. http://www.publicagenda.org/ • National Political Services, Inc. - provides survey research and strategic services specializing in public opinion research and strategic consulting services for political candidates, public ... http://www.nationalpolitical.com/polling.htm • Zogby International - Mission: "To Offer the Best Polling, Market Research, & Information Services Worldwide. Based on Accuracy & Detailed Strategic Information." http://www.zogby.com/ • The Brookings Institution is an independent, nonpartisan organization devoted to research, analysis, education, and publication focused on public policy issues in the areas of economics, foreign policy, and governance. http://www.brook.edu/ • Jack Burden & Associates - offers research, communication and strategic services to political candidates, nonprofit organizations, professional firms, and corporations. http://www.jackburden.com/ 24
25. 2. Companies that work for Republicans only • The Tarrance Group – uses opinion-based research to formulate strategic advice for political leaders, special interest groups, and major corporations. http://www.tarrance.com/ • Tel Opinion Research – was formed in 1995 as a national market research company that could provide high quality market research and polling at a fair and reasonable cost to clients in both the private and public sector. http://www.telopinion.com/ • The Winston Group - provides strategic survey research individually designed to meet the distinct needs of each client and characterized by its intellectual perspective and total objectivity. http://www.winstongroup.net/ • Public Opinion Strategies - the largest Republican polling firm in the country. http://www.pos.org/ • McLaughlin and Associates - specialize in public opinion research, media planning and buying services, and strategic consulting services. http://www.mclaughlinonline.com/ • American Viewpoint - can design and field a questionnaire that will obtain the data you need to successfully run for public office; pass a referendum or initiative; learn more about your association’s membership; influence legislation; test a marketing strategy; or launch a new product. http://www.amview.com/ • The Heritage Foundation - A leading conservative think tank, which publishes research on domestic, economic, foreign and defense policy, as well as The Journal Policy Review. http://www.heritage.org/ • Kolpien and Associates - political consulting and campaign services specializing in traditional Republican campaigns as well as consulting for non-partisan referendums and issue campaigns. http://www.kolpienassociates.com/ • Allan Hoffenblum & Associates - a full-service political consulting firm specializing in the management of Republican political campaigns. http://www.hoffenblum.com/ • Spett Strategies - provides strategic new and traditional media consultation and production services to Republican candidates and like-minded advocacy groups. http://www.spettstrategies.com/ 3. Companies that work for Democrats only • Peter D. Hart Research Associates - Its political division has worked with more than 40 United States senators and 30 governors and on more than 400 political campaigns, which has given the 25
+1-315-636-4485
+91 98151-74050 
